Crowding out or crowding in? The interaction between public pensions and private transfers in China

ABSTRACT

Public pensions and intergenerational family support are among the most important and well-researched income sources for older individuals in many societies. However, the interactions between them have received less attention. A few exceptions discuss if public transfers ‘crowd out’ private transfers. Yet, these studies in developed welfare states are considered misplaced due to their ‘fait accompli’ support arrangements. Furthermore, the cultural and symbolic value of support is often overlooked in these debates. Focusing on the case of China, where family support is institutionalized and public pension coverage is expanding, this study aims to contribute to the ‘crowding-out’ debates by examining intergenerational family transfers following the recent public pension reforms. Based on Heckman two-step regression analyses of the China Health and Retirement Longitudinal Study (2011–2015), the study shows that the ‘crowding-out’ and ‘crowding-in’ effects of public pensions co-exist. The long-established employee pension schemes reduced child-to-parent transfers, but increased parent-to-child transfers. More importantly, access to the new resident pensions, rather than displacing family support, has in effect ‘crowded in’ more child-to-parent transfers. These findings provide new insights into the dynamics between working-age and older generations, as well as between the state and family concerning their roles in welfare provision.

KEYWORDS:

• Public pension
• private transfer
• China
• family support
• ageing

1. Introduction

Intergenerational relationships, whether manifested at the macro-level such as public pension regulations or at the micro-level such as parent-child support, have been a focal point of research on ageing and public policy. Public transfers, particularly public pensions, constitute the most crucial source of income for older individuals in many societies, contributing more than 80% of the total incomes for those aged over 65 years in countries including Austria, Belgium, Finland, and Luxembourg (OECD, 2021). Meanwhile in societies with inadequate public pensions, family is often the immediate source of financial support for older people (Izuhara, 2010; Walker & Wong, 2005). Despite the importance of the two types of intergenerational transfers, the interactions between public pensions and family support have received less attention, especially in societies with limited pension coverage and extensive family support.

A few exceptions that examine such interactions focus on the underlying motives behind private transfer behaviours. Based on the altruism motive model, Becker (1974) claims that public transfers will displace private transfers within the families. Whereas Cox (1987) argues that private transfers may be ‘crowded in’ by public transfers under the exchange motive model. These theoretical debates inspired a series of subsequent empirical research in well-established welfare states. However, findings from these societies with large public transfers are considered misplaced due to their ‘fait accompli’ support arrangements (Cox et al., 2004, p. 2194). Moreover, critics of these approaches underline the importance of taking contextual factors into consideration. These factors include the mixed responsibility and division of labour of support between the state and family, the impact of cultural and social norms on family support arrangement, and the symbolic values of private transfers (Kohli, 1999; Motel-Klingebiel et al., 2005; e.g.; Blome et al., 2009). These developments highlight the needs to shift the focus away from established welfare states to societies with limited public transfers such as China and other East Asian societies. In addition, they also underline the importance of moving beyond transactional behaviours, as well as a consideration of the non-monetary value of intergenerational support.

Therefore, focusing on the case of China, where public pension transfers are limited and the role of families in old-age support is institutionalized (Abrahamson, 2017), this paper will complement the existing literature and provide valuable insights into the interactions between public pensions and family support. The paper will first review the theoretical debates and empirical evidence on the ‘crowding-out/in’ debate, highlighting the research gap. Then it will present the case of China by providing an overview of old-age financial support arrangements in both public and private spheres. Next, the paper will explain the datasets, measures, and analytical approaches used in this study. Following this, the paper will present its key findings which will focus on the impact of public pensions on both upward and downward private transfers. To conclude, the paper will discuss the implications of such interactions and potential avenues for future work.

2. ‘Crowding out’ or ‘Crowding in’: an ongoing debate

Traditionally, discussions regarding the interplay between public pensions and family support have centred on the two dominant motives underpinning private transfer behaviours within families. These are the altruism motive and the exchange motive. The altruism motive, as conceptualized within economic theory, posits that private transfers are driven by a concern for the utility of other family members (Becker, 1974; Becker & Barro, 1988). In the context of intergenerational private transfers, an adult child provides support for their older parents, with the understanding that the utility or wellbeing of the adult child is partially dependent on the wellbeing of their older parents, and vice versa. In practice, a pure altruism model, consistent with this framework, assumes that public transfers to older people increase their wellbeing as effectively as private transfers do so. Consequently, public provisions may replace, or ‘crowd out’ private transfers from adult children.

By contrast, the exchange motive offers an alternative framework for understanding private transfers. Cox (1987) argues that the wellbeing of an adult child is partially determined by the family support they received from their older parents such as care and services. In the event of public transfers, older individuals become financially empowered and consequently less dependent on family support. As Cox (1987) demonstrates, this leads to a decrease in the marginal utility of consumption. As a result, the adult child needs to provide more private transfers to achieve the same level of care and services from their older parents. This approach tends to view intergenerational support behaviours, either in the form of financial transfers or services, as ‘part of a transaction that contains a quid pro quo’ (Cox, 1987, p. 508), where individuals make rational choices to maximize their individual interests. This process is often described as the ‘crowding-in effect’, where public transfers stimulate more private transfers within families.

Earlier evidence indicates that the ‘crowding-out effect’ of public transfers may be of less importance in developed countries than in developing countries.¹ Some scholars point out that developed countries with large public transfers are the wrong place to look for the responsive transfer behaviours, as there may have been a ‘fait accompli’ support arrangement (Cox et al., 2004, p. 2194), where informal support mechanisms have been ‘gradually eroded’ by the long standing practice of public transfers (Nikolov & Bonci, 2020, p. 2). Therefore, the ideal place to examine such interactions is developing countries where public transfers are relatively small, and as a result, responsive transfers can be better detected. Moreover, the significance of private transfers in developing countries is more pronounced for its common roles as a primary safety net and risk-sharing mechanism (García & Cuartas, 2021). Responding to this call, some scholars have identified a significant ‘crowding-out effect’ in developing societies. For instance, in Peru, social security benefits reduce child-to-parent transfers by 20% (Cox & Jimenez, 1992); in rural Mexico, old-age public transfers lower the likelihood of receiving private transfers by 31% (Amuedo-Dorantes & Juarez, 2013); in South Africa, each additional unit of pension income for older individuals is associated with a 0.25–0.30-unit fewer private transfers (Jensen, 2004). A more recent review of studies in low- and middle-income countries also suggest that social protection programmes reduce the probability or amount of private transfers (Nikolov & Bonci, 2020).

However, it is argued that the altruism and exchange motives may co-exist and may respectively predominate in different parts of the income distribution (Cox et al., 2004, p. 2196). This hypothesis was substantiated by Maitra and Ray’s study (2003), which reveals that in South Africa the crowding-out effect holds only for low-income households. Consequently, the interactions between public pensions and private transfers, in fact, contributes to inequality due to their ‘asymmetric effects’ on households at different income levels (Maitra & Ray, 2003, p. 43). Similarly, research in South Korea shows that the crowding-out effect does not apply to all households but only for middle-income families (Lee & Lee, 2009). These findings, on the one hand, emphasize the potential heterogeneity in these interactions, whilst on the other, they underline the need to consider within-family transfers when examining the redistributive effects of public transfers (e.g. Bennett, 2013).

Evidence also suggests that the ‘crowding-in’ and ‘crowding-out’ effects co-exist when examining the bi-directional flow of intergenerational support. For instance, Reil-Held (2006) found that older individuals in Germany are positioned not merely as benefactors but also as beneficiaries of private transfers. Their study finds that public transfers to older people increase parent-to-child private transfers (i.e. crowd in downward private transfers) and decrease child-to-parent private transfers (i.e. crowd out upward private transfers). The ‘crowding-in’ of downward private transfers from older people to adult children provides evidence that public transfers to older generations are partially channelled back to younger generations via informal support networks.

Moreover, a series of comparative research studies have also challenged the dichotomy in support motives by bringing in institutional and contextual factors to the debates. One of the most influential publications adopting this approach, by Künemund and Rein (1999), compares family support in five countries with different welfare spending levels (United States, United Kingdom, Canada, Germany, and Japan) and identifies a positive relationship between welfare benefits and intergenerational support. Similar findings have also been reported in later comparative research studies (Blome et al., 2009; Brandt & Deindl, 2013; Motel-Klingebiel et al., 2005). These findings suggest that public transfers to older people may reduce potential conflicts over financial resources between generations (similar arguments in the context of Greater China can be found in Ko & Möhring, 2021; Peng et al., 2021) and therefore enhance family solidarity and create new forms of reciprocity. Moreover, they highlight the importance of the often neglected cultural and social norms on family support arrangements and the symbolic and emotional values of private transfers. This paper will therefore build upon the existing literature by relating these to the Chinese context.

3. The case of China

Unlike many developed welfare states in the West, in China and many East Asian societies, Confucian culture predominates and familistic characters of welfare provision are prominent. In China, family support behaviours are deeply influenced by social norms and have been institutionalized by social policies (Abrahamson, 2017; Jones, 1990; Walker & Wong, 2005). Despite this cultural and policy landscape, only a limited number of studies have attempted to explore the potential interaction between public pensions and private transfers in China. This is particularly surprising in light of the recent expansion of public pension coverage which began in 2009.

The origins of China’s pension system in the 1950s established a framework for ‘cradle-to-grave’ benefits, where only retired workers in government, state-owned enterprises, and public institutions were covered by public provisions (the later Government and Institution Pensions, hereafter, GIP). The transition from a planned economy to a market economy led to a shrinking size of the public sector and the emergence of a private and informal sector. To facilitate economic reform, public pension coverage was extended to employees in the private sector and later the self-employed (Enterprise Employee Basic Pensions, hereafter, EEBP). However, until 2009, rural farmers and the urban non-employed population had been excluded from the formal public pension system. For these groups, savings and private transfers from adult children were the predominant, if not only, available sources of income in old age.²

In 2009, a new pilot public pension scheme for rural residents (New Rural Social Pensions, hereafter, NRSP) was introduced. Shortly after, in 2011, a similar public pension scheme was initiated to further extend the coverage of public pensions to urban non-employed residents (Urban Resident Social Pensions, hereafter, URSP). The two new pension schemes (i.e. NRSP and URSP) were soon rolled out across China and then merged into a single scheme in 2014. The number of recipients of the new pensions saw a sharp increase from 16 million in 2009 to 131 million in 2012, and then to 161 million in 2020 (Ministry of Human Resources and Social Security of China, 2021). Together with the existing pension schemes for employees (i.e. GIP and EEBP), the new public pension arrangements aimed at a full coverage, with a fragmented and stratified structure reflecting occupational and urban-rural divisions.³

Although the new public pensions have drawn significant interest from researchers exploring the sustainability and adequacy of the public pension provisions (T. Liu & Sun, 2016; Tao, 2017; Williamson et al., 2017), very few studies have examined the potential response of intergenerational family support, especially in light of the primary role that the Chinese family has been playing in old-age support arrangements (Abrahamson, 2017; Izuhara, 2010; Walker & Wong, 2005). For instance, adult children providing support for older parents is not only in accordance with the code of conducts of Chinese filial piety and Confucian culture (Croll, 2001; Yeh et al., 2013) but also constituted as a legal obligation.⁴ Such support arrangement and resource exchanges within the family sphere also highlight the role of familial agency in protecting its members and achieving security, particularly within the political economy of East Asian welfare states (J. Liu, 2023; Papadopoulos & Roumpakis, 2017). Therefore, it is vital to understand how macro-level public transfers across generations (i.e. the Pay-As-You-Go public pension system) impact the patterns of micro-level private intergenerational transfers within the family sphere.

Previous studies on such interactions in China have shown inconsistent findings and highlighted urban-rural differences. Some evidence indicates that the crowding-out effect is insignificant for rural households (Gibson et al., 2011), minor or even reversed for urban households (T. Chen et al., 2017). However, other research focuses exclusively on the impact of social pensions on rural families, many of which used the China Health and Retirement Longitudinal Study. For example, Nikolov and Adelman (2019) find a ‘crowding-out effect’ of new rural pensions on the occurrence of private transfers, while both Wang and Chen (2017) and Ko and Möhring (2021) report a significant ‘crowding-in effect’ on the amount of private transfers received by older people in rural China. Conversely, qualitative studies in urban and rural China reported that intergenerational family support remains prevalent regardless of the pension status of older individuals (J. Liu, 2022, 2023). Beyond variations in research designs, analytical methodologies, and sample selection strategies, this inconsistency might also be attributed to the temporal lag between the introduction of new pensions and subsequent shifts in family support arrangements – a consideration particularly salient given the cultural underpinnings of Chinese intergenerational relationships. Moreover, these findings further reflect the urban-rural dichotomy within Chinese society, influencing welfare provision arrangements and familial support dynamics (Lau, 2013; Li & Piachaud, 2004; T. Liu & Sun, 2016; S. J. Shi, 2012).

In summary, the existing literature, coupled with recent policy shifts, highlight the need to examine the impact of public pensions on intergenerational private transfers within Chinese families. Furthermore, the distinctive context of China, where family support assumes a vital role, financial transfers possess cultural and symbolic significance, and public pension provision is experiencing unprecedented expansion, will offer new insights into the broader debates surrounding the ‘crowding-out/in’ effects.

4. Research aim and hypotheses

This study aims to understand the potential impact of public pensions on intergenerational private transfers within Chinese families. From previous studies, two themes have emerged. First, the potential impact of public pensions can be understood either by the participation in a specific pension scheme or by the amount of pension incomes received by older individuals (e.g. Nikolov & Bonci, 2020). Second, the intergenerational support flows bi-directionally (e.g. Grundy, 2005; Guo et al., 2012; Reil-Held, 2006). Considering these observations, the following null hypotheses are developed:

Hypothesis 1: Participation in public pension schemes does not impact the amount of upward intergenerational transfers received by older individuals.

Hypothesis 2: The amount of public pension income does not impact the amount of upward intergenerational transfers received by older individuals.

Hypothesis 3: Participation in public pension schemes does not impact the amount of downward intergenerational transfers provided to adult children.

Hypothesis 4: The amount of public pension income does not impact the amount of downward intergenerational transfers provided to adult children.

As discussed in the substantial body of literature (Lau, 2013; Li & Piachaud, 2004; T. Liu & Sun, 2016; S. J. Shi, 2012), long-standing social and economic disparities exist between urban and rural China. To further capture the possible heterogeneity, an additional hypothesis is included to examine whether such impact differs between urban and rural residents.

Hypothesis 5: There is no difference in the impact of public pensions on intergenerational transfers between urban and rural residents.

5. Research design

5.1. Data and measures

This study utilizes the China Health and Retirement Longitudinal Study (hereafter, CHARLS). CHARLS is recognized as one of the most comprehensive and high-quality datasets focusing on older population in China. This large-scale nationally-representative dataset has been employed in numerous empirical studies to explore social issues relevant to older individuals and their families (e.g. Ning et al., 2016; Yang et al., 2016; Zhao & Zhao, 2018). The national CHARLS baseline survey was carried out in 2011 and followed up every two to three years thereafter. This study analyses three waves of CHARLS data (2011, 2013 and 2015) spanning the rapid expansion of the new pension schemes for urban and rural residents. Respondents aged below 60 years at the time they were surveyed were excluded from the sample to allow for comparisons across gender, employment type, and pension schemes. Following this process, the final sample consisted of 14,207 observations with key information on public pensions schemes and 8,836 observations with key information on pension incomes.

The dependent variables are 1) the incidence and amount of upward private transfers received by older people from their adult children, and 2) the incidence and amount of downward private transfers provided by older people to their adult children. Respondents in the survey are asked to recall the amount of in-kind and monetary financial support received from/provided to non-co-residing adult children in the past year.⁵ The key explanatory variables include the specific public pension schemes from which respondents claimed pensions, reflecting the fragmented and stratified nature of public pension system in China, and the logarithmic value of the public pension benefits that the respondents received.⁶ Both of these measures have been extensively discussed in previous literature (X. Chen et al., 2018; Cox et al., 2004; Jensen, 2004; Nikolov & Adelman, 2019). A binary variable, hukou (i.e. urban or rural resident), is used as a moderator variable aiming to capture the potential urban-rural heterogeneity in the interactions between public pensions and private transfers. Informed by previous literature (Cox et al., 2006; Reil-Held, 2006; e.g.; Amuedo-Dorantes & Juarez, 2013; Brandt & Deindl, 2013; T. Chen et al., 2017; X. Chen et al., 2018; Nikolov & Adelman, 2019; Nikolov & Bonci, 2020), a set of control variables have been incorporated. These include the pre-transfer income levels of the respondents, socio-demographic characteristics of respondents and of their adult children, as well as variables related to intergenerational relationships. A full list and a brief descriptive summary of the control variables can be found in Supplementary Material.

5.2. Analytical approaches

The study applies the Heckman two-step selection model for panel data with random effects. The primary reason for using this two-step selection model is to reduce potential selection bias. For instance, if there were systematic differences between respondents who did receive upward private transfers (selected samples) and those who did not (unselected samples), then the estimation of the impact of public pensions on the amount of upward private transfers, based on the selected samples, would be inaccurate. The other reason is that both extensive margin (incidence) and intensive margin (amount) are worth examining to fully understand the possible impact of public pensions, as private transfers are decided in two stages (Becker, 1974; Nikolov & Adelman, 2019). Similar analytical approaches have also been used in previous literature (e.g. Cox, 1987; Reil-Held, 2006).

The models are as follow:

$\left.Y_{Fa\mathrm{mily}su\mathrm{pport}}^{it}={\beta }_1\mathit{Public}\mathit{pensio}{n}_{\mathit{it}}+{\beta }_2\mathit{Urban}\mathit{huko}{u}_{\mathit{it}}+{\beta }_3\mathit{Public}\mathit{pension}\ast \mathit{Urban}\mathit{huko}{u}_{\mathit{it}}+{\beta }_{4-10}\mathit{Respondent}\mathit{related}\mathit{variable}{s}_{\mathit{it}}+{\beta }_{11-15}\mathit{Child}\mathit{related}\mathit{variable}{s}_{\mathit{it}}+\mathit{\gamma E}(S_{\mathit{it}}=1|S_{\mathit{it}}\right)+{\alpha }_i+{\epsilon }_{\mathit{it}}$

where,

$\begin{array}{rl}{S}_{\mathit{it}}={\delta }_1\mathit{Public}& \mathit{pensio}{n}_{\mathit{it}}+{\delta }_2\mathit{Urban}\mathit{huko}{u}_{\mathit{it}}\\ & +{\delta }_3\mathit{Public}\mathit{pension}\\ & \ast \mathit{Urban}\mathit{huko}{u}_{\mathit{it}}+{\delta }_{4-9}\mathit{Respondent}\mathit{related}\mathit{variables}\\ & +{\delta }_{10-14}\mathit{Child}\mathit{related}\mathit{variables}+v_i+u_{\mathit{it}}\end{array}$

In the above models, $Y_{Fa\mathrm{milysupport}}^{it}$ represents the (logarithmic) value of private transfers received or provided by older people, and $\mathit{Publicpensio}{n}_{\mathit{it}}$ represents either a specific public pension scheme or the (logarithmic) pension incomes. $S_{\mathit{it}}$ represents the probability of the respondent receive or provide private transfers from or to their adult children.⁷

6. Upward child-to-parent private transfers

Table 1 reveals that a majority of respondents (71.71%) received upward private transfers from their adult children in the past 12 months. The prevalence of upward private transfers was found to be highest among NRSP/URSP recipients (81.63%) and lowest among recipients of the two long-established public pension schemes (58.13% for GIP and 54.44% for EEBP, respectively). However, GIP and EEBP recipients on average received more upward private transfers compared to other groups (Table 2). Additionally, there is a statistically significant yet weak positive correlation between the logarithmic values of pension incomes and upward private transfers received by respondents.

Table 1. Incidence of receiving upward private transfers by public pension schemes.

Whether older people received upward private transfers		Total	Recipients of GIP	Recipients of EEBP	Recipients of NRSP/URSP	Not in any public pensions
	Yes	10188	558	914	5603	3113
		(71.71%)	(58.13%)	(54.44%)	(81.63)	(66.16%)
	No	4020	402	765	1261	1592
		(28.29%)	(41.88%)	(45.56%)	(18.37%)	(33.84%)

Pearson chi2(3) = 738.4160 Pr = 0.0000

NB: First row has frequencies and second row has column percentages in parentheses.

Data source: CHARLS 2011, 2013 and 2015

Table 2. Associations between logarithmic private transfers received by older people and public pensions.

	Public pension schemes
Private transfers (log) received by older people	Recipients of GIP	Recipients of EEBP	Recipients of NRSP/URSP	Not in any public pensions
	(n = 558)	(n = 914)	(n = 5603)	(n = 3113)
	8.213	8.156	7.755	7.621
	F-test: Ho: mean (None) = mean (GIP) = mean (EEBP) = mean (NRSP/URSP)				Pr = 0.0000
		Logarithmic pension incomes (n = 6517)
	0.122
	Pearson‘s correlation test: Ho: ρ = 0				Pr = 0.0000

NB: Only 6517 out of 8836 respondents reported non-zero pension incomes and upward private transfers.

Data source: CHARLS 2011, 2013 and 2015.

To delve deeper into the impact of public pensions subsequent regression models were presented in Table 3. Model 1 uses public pension schemes as the key explanatory variable, whilst, Model 2 employs logarithmic pension incomes as the key explanatory variable. Within each model, two equations are featured: a selection equation (Equation 2) to estimate the probability of the occurrence of upward private transfers to older people, and an adjusted equation (Equation 1) to estimate the impacts of public pensions on the logarithmic amount of upward private transfers to older individuals, providing a correction for potential sample selection issues.⁸

Table 3. Heckman two-step selection panel model estimating the impacts of public pensions on upward private transfers received by older people.

Upward private transfers received by older people (logarithmic)
	Model 1 Pension schemes on received transfers		Model 2 Pension incomes on received transfers
	(1)	(2)	(1)	(2)
	Adjusted Equation	Selection Equation	Adjusted Equation	Selection Equation
Urban hukou	0.127*	−0.216***	0.367	−0.337
	(0.069)	(0.072)	(0.351)	(0.386)
Pension schemes (ref. Not in any public pension schemes)
GIP	0.169	−0.012
	(0.138)	(0.154)
EEBP	0.143	−0.311**
	(0.139)	(0.134)
NRSP/URSP	0.117***	0.157***
	(0.034)	(0.039)
Pension schemes *urban hukou (ref. Not in any public pension * Rural hukou)
GIP*urban hukou	−0.168	−0.122
	(0.162)	(0.175)
EEBP*urban hukou	−0.043	0.147
	(0.156)	(0.150)
URSP*urban hukou	−0.206**	−0.080
	(0.105)	(0.119)
Pension incomes(log)			−0.010	−0.140***
			(0.029)	(0.032)
Pension incomes(log) *urban hukou			−0.031	0.019
			(0.041)	(0.044)
Pre-transfer Incomes level (ref. incomes below p25)
Incomes between p25 and p50	−0.057	0.121***	−0.051	0.173**
	(0.037)	(0.044)	(0.065)	(0.079)
Incomes between p50 and p75	−0.149***	0.072*	−0.157**	0.157**
	(0.038)	(0.043)	(0.067)	(0.079)
Incomes above p75	0.018	−0.054	−0.005	0.047
	(0.050)	(0.053)	(0.079)	(0.087)
Age		0.007***		0.007*
		(0.003)		(0.004)
Female	0.117***	0.190***	0.134***	0.185***
	(0.033)	(0.036)	(0.039)	(0.046)
Education background (ref. Illiterate)
Less than primary school	0.056	0.071	0.056	0.053
	(0.041)	(0.046)	(0.050)	(0.062)
Primary school	0.085**	0.098**	0.114**	0.121*
	(0.042)	(0.046)	(0.050)	(0.062)
Secondary school	0.279***	0.012	0.301***	0.043
	(0.053)	(0.057)	(0.063)	(0.075)
High school	0.411***	−0.098	0.472***	0.043
	(0.073)	(0.072)	(0.082)	(0.088)
Marital status of respondent (ref. Married and living with spouse)
Married but temporarily living alone	−0.069	0.073	−0.034	−0.004
	(0.074)	(0.083)	(0.095)	(0.115)
Widowed/separated/divorced	−0.171***	−0.040	−0.216***	−0.030
	(0.034)	(0.039)	(0.040)	(0.051)
Self-assessed health (ref. Very good)
Good	0.003	0.060	0.010	0.062
	(0.059)	(0.067)	(0.073)	(0.089)
Fair	−0.044	0.107*	−0.080	0.059
	(0.051)	(0.058)	(0.062)	(0.076)
Poor	−0.080	0.085	−0.130**	0.010
	(0.054)	(0.060)	(0.065)	(0.080)
Province by economic development level (ref. High level)
Middle-level	−0.209***	0.015	−0.196***	0.049
	(0.039)	(0.041)	(0.047)	(0.054)
Low-level	−0.130***	0.054	−0.084*	0.086
	(0.038)	(0.041)	(0.046)	(0.054)
N. of children	0.201***	0.290***	0.190***	0.282***
	(0.011)	(0.013)	(0.013)	(0.018)
Co-residence	−0.109***	−0.185***	−0.002	−0.186***
	(0.034)	(0.037)	(0.043)	(0.051)
Married children	0.099	0.321***	0.045	0.367***
	(0.061)	(0.060)	(0.076)	(0.079)
Proximity to adult children (ref. Same community)
Same city	−0.135**	−0.058	0.002	−0.044
	(0.058)	(0.055)	(0.073)	(0.073)
Same province	0.094	0.446***	0.128*	0.535***
	(0.060)	(0.061)	(0.073)	(0.079)
Different province	0.579***	0.491***	0.622***	0.576***
	(0.059)	(0.063)	(0.073)	(0.084)
Incomes of children (ref. Incomes less than 10k)
10k-20k	−0.028	−0.005	−0.089	0.012
	(0.049)	(0.051)	(0.062)	(0.070)
20k-50k	0.308***	0.241***	0.278***	0.178***
	(0.041)	(0.045)	(0.050)	(0.059)
50k+	0.891***	0.207***	0.878***	0.277***
	(0.054)	(0.060)	(0.064)	(0.076)
Frequency of contact (ref. Very frequent)
Frequent	−0.299***	0.096*	−0.246***	0.013
	(0.045)	(0.054)	(0.054)	(0.070)
Infrequent	−0.856***	−1.753***	−0.784***	−1.878***
	(0.135)	(0.117)	(0.168)	(0.157)
Very infrequent	−0.671***	−0.370***	−0.713***	−0.521***
	(0.055)	(0.054)	(0.070)	(0.075)
2013	−0.198***	0.698***	−0.304***	0.893***
	(0.042)	(0.043)	(0.060)	(0.060)
2015	0.148***	1.118***	0.002	1.292***
	(0.045)	(0.046)	(0.063)	(0.067)
Var(e.amount)	1.163***		1.055***
	(0.029)		(0.039)
Corr(e.receive,e.amount)	−0.210***		−0.229***
	(0.056)		(0.073)
Var(amount[id])	0.589***		0.598***
	(0.033)		(0.045)
Var(receive[id])	0.396***		0.409***
	(0.049)		(0.077)
Corr(receive[id],amount[id])	0.532***		0.660***
	(0.061)		(0.090)
Constant	6.918***	−1.920***	7.245***	−0.961***
	(0.123)	(0.197)	(0.256)	(0.352)
Observations	14,207		8,836
Selected	10188		6517
Non-selected	4019		2319
Number of id	8,152		6,057

Standard errors in parentheses

*** p < 0.01, ** p < 0.05, * p < 0.1

NB:Var(e.amount) is an estimate of the variance of the observation-level error for amount of upward private transfers. Corr(e.receive,e.amount) is an estimate of its correlation with the observation-level error for the selection model. Var(receive[id]) is an estimate of the variance of the random effects. Corr(receive[id],amount[id]) is an estimate of the correlation of the random effects.

The results detailed in Table 3 depict a complex interaction between public pensions and upward private transfers, revealing both ‘crowding out’ and ‘crowding in’ effects. The selection equation in Model 2 shows that in general, the probability of receiving upward private transfers decreased with pension incomes. Specifically, ceteris paribus, a 20% increase in pension incomes would reduce the probability of older individuals receiving upward private transfers by 0.01 in its z-score. However, this ‘crowding-out’ effect was found to be statistically significant solely among EEBP recipients. As indicated in Model 1, access to EEBP benefits reduced the probability of receiving upward private transfers by 0.31 in its z-score.

Conversely, the newly introduced pensions, NRSP and URSP, increased the probability of receiving upward private transfers by 0.16 in its z-score, illustrating a ‘crowding-in’ effect. It is worth noting that the new pensions for rural residents not only increased the likelihood of receiving private transfers but also ‘crowd in’ the amount of private transfers by 12.41%. In contrast, among urban residents, the new pensions reduced the amount of private transfers by 8.51%.

Table 3 also presents the impact of other socio-demographic factors on upward private transfers. Taking Model 1 as an example, the probability for urban residents to receive upward private transfers was 0.22 lower in its z-score compared to their rural peers. However, on the condition that upward private transfers occurred, urban residents received 13.54% more private transfers than their rural counterparts. Older individuals with moderate incomes (i.e. where pre-transfer incomes fall within the 2nd to the 3rd quartile range) were more likely to receive upward private transfers, but the amount was lower compared to their counterparts. The probability of receiving upward private transfers increased with age. Gender difference was also significant, with women more likely to receive upward private transfers and the amount being greater. Education, in general, played a positive role in upward private transfers. Older individuals who were widowed, separated, or divorced received fewer transfers than married older individuals. Regional differences in the amount of upward private transfers were also statistically significant: those residing in less developed provinces received fewer private transfers than their peers in developed provinces. Self-assessed health conditions of older people, however, had no impact on the amount of upward private transfers. For them, care resources might be in more urgent need than monetary transfers.

Regarding intergenerational factors, both the incidence and amount of upward private transfers increased with the number of adult children. Living with at least one adult child decreased the probability of receiving private transfers, and the amount was also lower. In addition, having married adult children increased the probability of upward private transfers. For instance, the probability of receiving upward private transfers for older people whose children were all married was 0.32 higher in its z-score compared to their peers whose children are not married. Both the incidence and amount of upward private transfers received by older individuals were significantly higher if adult children have moderate to high incomes (more than 20k yuan per year). Interestingly, geographic distance stimulated a higher probability and more private transfers, but infrequent intergenerational contacts lowered the probability and reduced the amount of private transfers.

7. Downward parent-to-child private transfers

Access to public pensions improves the economic situation of older individuals. This economic empowerment may impact intergenerational support by enabling Chinese parents to continue to act as the ‘provider’ even in their old age. This section investigates the potential impact of public pensions on downward private transfers from older parents to adult children. Compared to upward private transfers, the incidence of downward private transfers among Chinese older people was lower. Specifically, only 30% of respondents provided private transfers to their adult children in the past year, as indicated in Table 4. Both the incidence and amount of downward private transfers were found to be highest among GIP and EEBP recipients. This was followed by NRSP/URSP recipients and older individuals without public pensions, as detailed in Tables 4 and 5. Furthermore, there was a moderately positive correlation between the amount of downward private transfers provided by older individuals and the amount of public pension incomes they received.

Table 4. Incidence of providing downward private transfers by public pension schemes.

Whether older people provided downward private transfers		Total	Recipients of GIP	Recipients of EEBP	Recipients of NRSP/URSP	Not in any public pensions
	Yes	4300	432	712	2088	1068
		(30.26%)	(45.00%)	(42.41%)	(30.42%)	(22.70%)
	No	9908	528	967	4776	3637
		(69.74%)	(55.00%)	(57.59%)	(69.58%)	(77.30%)

Pearson chi2(3) = 343.7146 Pr = 0.000

First row has frequencies and second row has column percentages in parentheses

Data source: CHARLS 2011, 2013 and 2015.

Table 5. Associations between logarithmic private transfers provided by older people and public pensions.

Private transfers (log) provided by older people	Public pension schemes
	Recipients of GIP	Recipients of EEBP	Recipients of NRSP/URSP	Not in any public pensions
	(n = 432)	(n = 712)	(n = 2088)	(n = 1068)
	8.281	7.942	6.681	6.749
	F-test: Ho: mean (None) = mean (GIP) = mean (EEBP) = mean (NRSP/URSP)				Pr = 0.0000
	Logarithmic pension incomes (n = 2959)
0.412
	Pearson‘s correlation test: Ho: ρ = 0			Pr = 0.0000

NB: Only 2959 out of 8836 respondents reported non-zero pension incomes and downward private transfers.

Data source: CHARLS 2011, 2013 and 2015.

The Heckman two-step selection estimations⁹ in Table 6 shows a positive but limited impact of public pensions on the occurrence of downward private transfers. As shown in Model 3, only EEBP increased the probability of providing downward private transfers by 0.237 in its z-score. The impact of other public pension schemes (i.e. GIP, NRSP/URSP) was not found statistically significant. A modest positive impact was also found when examining the impact of pension incomes (Model 4): for instance, a 20% increase in pension income only increased the probability of providing downward private transfers by 0.005 in its z-score, and 1.95% higher in amount. Urban residents were more likely to provide downward private transfers (0.12 higher in its z-score), and the amount was on average 91.36% higher compared to their rural counterparts. The interaction terms between residence type and public pensions or logarithmic pension incomes are not statistically significant, indicating that, all other factors being equal, urban and rural residents would experience the ‘crowding-in’ effect of public pensions in similar ways.

Table 6. Heckman two-step selection panel model estimating the impacts of public pensions on downward private transfers provided by older people.

Downward private transfers provided by older people (logarithmic)
	Model 3 Pension schemes on provided transfers		Model 4 Pension incomes on provided transfers
	(1)	(2)	(1)	(2)
	Adjusted Equation	Selection Equation	Adjusted Equation	Selection Equation
Urban hukou	0.649***	0.120*	−0.052	0.267
	(0.127)	(0.064)	(0.593)	(0.321)
Pension schemes (ref. Not in any public pension schemes)
GIP	0.337	0.072
	(0.252)	(0.131)
EEBP	0.234	0.237**
	(0.214)	(0.118)
NRSP/URSP	−0.028	0.037
	(0.072)	(0.035)
Pension schemes *urban hukou (ref. Not in any public pension * Rural hukou)
GIP*urban hukou	−0.220	0.141
	(0.284)	(0.151)
EEBP*urban hukou	−0.320	−0.081
	(0.242)	(0.132)
URSP*urban hukou	0.000	−0.056
	(0.193)	(0.100)
Pension incomes(log)			0.106**	0.060**
			(0.049)	(0.027)
Pension incomes(log) *urban hukou			0.041	−0.024
			(0.067)	(0.037)
Pre-transfer incomes level (ref. incomes below p25)
Incomes between p25 and p50	−0.122	0.010	−0.077	0.095
	(0.082)	(0.039)	(0.144)	(0.070)
Incomes between p50 and p75	−0.063	0.159***	−0.049	0.240***
	(0.084)	(0.039)	(0.148)	(0.071)
Incomes above p75	0.378***	0.239***	0.305*	0.332***
	(0.094)	(0.046)	(0.160)	(0.077)
Age		−0.016***		−0.018***
		(0.002)		(0.003)
Female	−0.027	0.044	−0.009	0.023
	(0.059)	(0.030)	(0.069)	(0.038)
Education background (ref. Illiterate)
Less than primary school	−0.006	0.069*	−0.041	0.069
	(0.079)	(0.039)	(0.097)	(0.050)
Primary school	0.162**	0.093**	0.197**	0.112**
	(0.079)	(0.039)	(0.095)	(0.049)
Secondary school	0.274***	0.111**	0.299***	0.152**
	(0.093)	(0.048)	(0.112)	(0.060)
High school	0.622***	0.209***	0.571***	0.282***
	(0.115)	(0.061)	(0.135)	(0.074)
Marital status of respondent (ref. Married and living with spouse)
Married but temporarily living alone	0.021	0.111	0.073	0.134
	(0.130)	(0.070)	(0.160)	(0.093)
Widowed/separated/divorced	−0.154**	−0.200***	−0.177**	−0.159***
	(0.073)	(0.033)	(0.083)	(0.041)
Self-assessed health (ref. Very good)
Good	−0.254**	−0.009	−0.271**	−0.013
	(0.107)	(0.057)	(0.129)	(0.072)
Fair	−0.230**	−0.022	−0.271**	−0.018
	(0.090)	(0.048)	(0.109)	(0.062)
Poor	−0.249**	−0.177***	−0.289**	−0.174***
	(0.100)	(0.051)	(0.120)	(0.065)
Province by economic development level (ref. High level)
Middle-level	−0.013	0.108***	0.012	0.107**
	(0.070)	(0.035)	(0.084)	(0.044)
Low-level	−0.116	0.128***	−0.095	0.135***
	(0.071)	(0.035)	(0.085)	(0.044)
N. of children	−0.027	0.051***	−0.043*	0.066***
	(0.019)	(0.010)	(0.023)	(0.013)
Co-residence	−0.038	−0.063*	0.078	−0.047
	(0.068)	(0.033)	(0.083)	(0.043)
Married children	−0.287***	−0.059	−0.389***	−0.059
	(0.107)	(0.053)	(0.129)	(0.068)
Proximity to adult children (ref. Same community)
Same city	−0.019	0.021	0.066	−0.013
	(0.112)	(0.051)	(0.137)	(0.066)
Same province	0.065	0.291***	0.139	0.308***
	(0.112)	(0.052)	(0.134)	(0.066)
Different province	0.171	0.170***	0.177	0.189***
	(0.111)	(0.053)	(0.134)	(0.068)
Incomes of children (ref. Incomes less than 10k)
10k-20k	−0.178*	−0.005	−0.226*	0.019
	(0.100)	(0.047)	(0.120)	(0.061)
20k-50k	0.237***	0.193***	0.128	0.173***
	(0.082)	(0.038)	(0.097)	(0.049)
50k+	0.619***	0.227***	0.511***	0.192***
	(0.097)	(0.049)	(0.112)	(0.061)
Frequency of contact (ref. Very frequent)
Frequent	−0.333***	−0.004	−0.283***	0.033
	(0.083)	(0.043)	(0.099)	(0.055)
Infrequent	−0.396	−1.049***	−0.225	−0.930***
	(0.322)	(0.116)	(0.367)	(0.148)
Very infrequent	−0.372***	−0.375***	−0.397**	−0.424***
	(0.129)	(0.052)	(0.162)	(0.068)
2013	−0.003	0.500***	−0.158	0.485***
	(0.102)	(0.039)	(0.123)	(0.051)
2015	0.390***	0.490***	0.140	0.367***
	(0.100)	(0.039)	(0.113)	(0.051)
var(e.amount)	1.994***		2.056***
	(0.129)		(0.172)
corr(e.provide,e.amount)	−0.256*		−0.311**
	(0.133)		(0.152)
var(amount [id])	0.609***		0.477***
	(0.096)		(0.120)
var(provide[id])	0.177***		0.232***
	(0.030)		(0.047)
corr(provide[id],amount[id])	0.336**		0.378*
	(0.141)		(0.198)
Constant	7.284***	−0.294*	6.900***	−0.639**
	(0.371)	(0.168)	(0.617)	(0.296)
Observations	14,207		8,836
Selected	4300		2959
Non-selected	9907		5877
Number of id	8,152		6,057

Standard errors in parentheses

*** p < 0.01, ** p < 0.05, * p < 0.1

Other factors that could affect downward private transfers are also presented in Table 6. Individuals whose incomes were above the median were more likely to provide downward private transfers to adult children. But only those in the highest income quartile provided significantly more transfers. Meanwhile, the probability of providing downward private transfers decreased with age. There was no significant evidence of gender differences. Both the incidence and amount of downward private transfers increased with the education level of older individuals. Widowed, separated, or divorced respondents were less likely to provide downward private transfers, and even if they did, the amount was smaller compared to married older individuals. Health condition was also significantly associated with both the incidence and amount of downward private transfers, especially for those who reported poor health. Overall, older people in more developed provinces were less likely to provide downward private transfers.

Moreover, the probability of providing downward private transfers increased with the number of adult children but decreased if the respondent co resided with at least one adult child. The amount of downward private transfers decreased if adult children are married. For example, those whose children were all married provided 24.95% fewer downward transfers than their peers whose children were all unmarried. Both the probability and amount of downward transfers were higher if adult children have moderate to higher incomes (more than 20k yuan per year). This seemingly counter-intuitive pattern demonstrates the inadequacy of viewing private transfers as merely functional and utility-driven, as it suggests even if adult children have more financial resources, they may still receive financial transfers from their older parents. Frequent contact between generations led to more downward private transfers provided by older individuals, however geographic distance, like its role in upward transfers, increased the likelihood of providing downward private transfers.

8. Discussion

The study reveals a nuanced picture in which the ‘crowding-out’ and ‘crowding-in’ effects of public pensions co-exist. The established public pensions, especially EEBP, have reduced the amount of child-to-parent upward financial transfers. This finding resonates with some earlier evidence from both developing countries like Peru (Cox & Jimenez, 1992), South Africa (Jensen, 2004), and Mexico (Amuedo-Dorantes & Juarez, 2013) as well as developed societies such as Germany (Reil-Held, 2006) and certain European countries (Deindl & Brandt, 2011). This observation can largely be understood through the altruism model (Becker, 1974; Becker & Barro, 1988), as frequently referenced and discussed in the above literature.

However, the impact of public pensions becomes more complicated when the ‘crowding-in’ effect of the new pensions (i.e. NRSP&URSP) is considered. Instead of lowering the likelihood of receiving private transfers from adult children, the new pensions, in fact, brought in more intergenerational family support. There are several possible explanations emerged from literature. The first, and perhaps most-discussed, is the exchange motive behind private transfers which requires the adult child to provide additional upward transfers in order to maintain the previous level of downward support (Cox, 1987). Another explanation could be that public pensions have empowered the previously marginalized older residents and enhanced their subjective wellbeing as well as alleviated financial stress within intergenerational families. This, in turn, may foster greater intergenerational solidarity and a more robust support network (Blome et al., 2009; Ko & Möhring, 2021; Peng et al., 2021). Moreover, within the family support network, intergenerational private transfers may carry non-monetary emotional and symbolic value, contributing towards the realization of the cultural tradition of filial piety, which profoundly shapes intergenerational relations and transfers in Chinese families (Izuhara, 2010; J. Liu, 2023; Zhong & Li, 2017). The persistence of intergenerational support arrangements may also reflect the inadequacy of the new pension benefits and demonstrate the resilience of Chinese families in exercising its agencies to protect its members (J. Liu, 2023; Papadopoulos & Roumpakis, 2017).

The empirical findings of this study further extend to the ‘crowding-in’ effect of public pensions on parent-to-child downward financial transfers, though this effect is found statistically significant solely among EEBP recipients. This study finds that public transfers to older individuals enhanced their active engagement in intergenerational support. This economic empowerment also illustrates that public resources given to older individuals were, at least partially, channelled back to younger generations through private transfers and intergenerational support networks within families. This observation provides new perspectives to the ongoing ideological and policy discourse concerning generational equity, interdependence, and conflict (Lau, 2013; Quadagno, 1989; Williamson et al., 2003), highlighting that younger generations may also benefit from the improvement of older individuals’ financial wellbeing. Furthermore, the noted heterogeneity in responsive transfer behaviours underlines the importance of accounting for intra-family transfers (Bennett, 2013) and redistributions in evaluating the redistributive effects of public and social policies. This consideration is particularly crucial for achieving equality across and within generations via policy interventions (Zhu & Walker, 2018).

In addition to the impact of public pensions, this study also discusses intergenerational support with reference to other socio-economic factors. Intergenerational private transfers are highly associated with the financial resources owned by both generations, but such relationships are not linear. Instead, the study suggests that the decision-making process for intergenerational financial transfers, whether upwards to older parents or downwards to adult children, may not be solely predicated on the financial needs of the recipient. This seemingly counter-intuitive finding highlights that intergenerational transfers transcend their functional value (e.g. see Daly, 2017; Summers, 2018) and encapsulate symbolic meanings (Grundy, 2005) such as compliance with legal requirements or the exemption from moral condemnations, particularly when adult children and their older parents reside separately. This dynamic is also mirrored in the interplay among intergenerational transfers, geographic proximity, and living arrangements between older individuals and their adult children. For instance, a greater geographic distance between older parents and their adult children leads to more and a greater likelihood of private transfers, aligning with the support arrangements in many intergenerational families with migrant workers (Cong & Silverstein, 2011). This study further reveals the continued involvement of older individuals in providing financial support to their adult children. It highlights the prolonged dependency of adult children on their parents and the evolving norms of filial piety, an emerging theme explored not only within the Chinese context but across other East Asian societies as well (Izuhara, 2010; J. Liu, 2023; Zhong & Li, 2017).

This study, focusing on the recent public pension reforms in China, presents a critical case for further understanding the dynamics between the state and family concerning their roles in welfare provision in East Asian societies. This exploration is particularly crucial considering the similarities in cultural and demographic characteristics as well as social policy arrangements, including the increasing involvement of the state in expanding institutional protection (Cheng et al., 2020). Through an analysis of the interplay between public pension systems and family transfers in China, the research demonstrates the assumed primary role of East Asian (intergenerational) families in supporting its older members, highlighting both the resilience of these practices and the enduring nature of familial ideologies (Abrahamson, 2017). Further research on these interactions is essential to understand how such arrangements response to ageing populations and changing family structures within a broader policy landscape.

This paper contributes to the ongoing ‘crowding-out/in’ debates by revealing the nuanced interplay between various support motives in different types of intergenerational support relationships. Moreover, they highlight the often-overlooked cultural and symbolic value of financial transfers within familial relationships. However, the current study is limited by the nature of its quantitative methodological assumptions that view measurable transfer behaviour outcomes as social reality and inevitably leave out individual and familial agency in the process of sensemaking and negotiation of these situations. The latter is especially important in understanding the ‘black box’ of financial redistributions within the family sphere (Bennett, 2013, p. 582; Pahl, 1989, p. 4). Therefore further in-depth (qualitative) studies are required to recognize, identify, and comprehend the contexts, mechanisms, and processes where family dynamics may react differently. Such understandings will enable a more profound insight into the complex relationships not only between the working-age and older generations, but also between the state and family concerning their roles in welfare provision in China and wider East Asian societies. Additionally, given the possible time lag in behavioural responses to policy changes – especially those influenced by entrenched social and cultural norms – incorporating a longer period of longitudinal observations could yield a more robust analysis.

9. Conclusion

This paper has examined the interactions between two levels of intergenerational transfers in the Chinese context where a fully-covered public pension system is under development and family support maintains a primary role in old-age support. While there exists a broad global literature examining the potential relationship between public pensions and family support (Arrondel & Masson, 2006; Brandt & Deindl, 2013; Deindl & Brandt, 2011; Motel-Klingebiel et al., 2005; Reil-Held, 2006), there has been limited engagement in the Chinese context. This study, therefore, contributes to the expanding research on the ‘crowding-out’ debates, as well as the diverse motives underlying intergenerational support.

Unlike previous research which has often overlooked the agency and initiative of older individuals, this study has acknowledged their status as both recipients and providers within intergenerational support networks (Arrondel & Masson, 2006; Deindl & Brandt, 2011; Reil-Held, 2006). The findings reveal that ‘crowding-out’ and ‘crowding-in’ effects co-exist, indicating that public pensions do not necessarily weaken intergenerational support, but rather, in some cases, strengthen these support networks. In addition, this paper also offers new insights into the intergenerational dynamics by highlighting the cultural and symbolic value of family transfers.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Supplemental data

Supplemental data for this article can be accessed online at https://doi.org/10.1080/17516234.2024.2367886.

Additional information

Notes on contributors

Jiaxin Liu

Jiaxin Liu is a Lecturer in Social and Public Policy at the University of York. Jiaxin’s research interests focus on social policy analysis and ageing. Specifically, Jiaxin is interested in public pensions, intergenerational relationships, family support, social policy in China/East Asia, and mixed-method research designs.

Notes

1. For instance, see Amuedo-Dorantes and Juarez (2013), Cox and Jakubson (1995), and Jensen (2004).

2. Due to the combination of a lack of financial literacy and trust, underdevelopment of pension products market, and regulations and policy, private or commercial pension plans were far less common among older people in China (Niu et al., 2020).

3. One of the most significant features of pension systems in East Asian societies is the higher level of fragmentation and stratification within the public pension structure, which often follows occupational lines and sometimes urban-rural divisions (Cheng et al., 2020; S.-J. Shi & Mok, 2012; Zhu & Walker, 2018).

4. For instance, The Law of the People’s Republic of China on Protection of the Rights and Interests of the Older (1996) is the first legislation focusing on older people in China, which says that it is the legal obligations of an adult child to ‘provide financial supports, daily life care, emotional supports, and meet special needs of older people’ (Chapter Two Article 11, 1996). In the 2012 revision, the Article ‘family should take the predominant responsibility to care and support for older people’ (Chapter Two Article 10, 1996) was changed into ‘the provision for older people should be based on family’ (Chapter Two Article 13, 2012). However, it still demonstrates the primary role of Chinese family in old-age support arrangements.

5. In a few cases, respondents may provide a range rather than a specific number. In this case, the mean values of the given range would be used as the proximity. It is also worth noting that the amount of upward/downward private transfers was transformed into their logarithmic forms in the analysis. Reasons for that were two-fold. From the practical aspect, the marginal utility of incomes declines as income increases (e.g. see Layard et al., 2008). For example, a one-thousand yuan increase in income not necessarily means the same for low-income older people as it means for high-income older people. From the statistical aspect, the distribution of upward/downward private transfers is highly skewed (the skewness coefficients far from 0). The logarithmic transformation, therefore, would reduce the skewness and increase its proximity to a normal distribution.

6. Both private transfer incomes and public pension incomes are inflation-adjusted with reference to 2015 price index.

7. The explanatory variables for the lower equation is a subset of the explanatory variables for the upper equation, and the upper has at least one explanatory variable that might affect the probability of being selected (the incidence of private transfers) but not the amount of transfers (as suggested by Semykina & Wooldridge, 2010). In this case, age is assumed to have an impact only on the probability of receiving upward transfers but not on the amount of upward transfers received by older people.

8. As reported in the bottom of Table 3, selection was an issue as the correlation terms (i.e. corr(e.receive,e.amount) and corr(receive[id],amount[id]), (as in Table 3) in both models were significant. Furthermore, taking Model 1 as an example, the estimate of correlation for the selection model is −0.201, which means that unobserved individual-level factors that increase the probability of receiving upward private transfers tend to decrease the amount of upward private transfers. On contrary, the estimate of correlation of random effects is 0.542, which means that unobserved observation-level (time-varying) factors that increase the probability of receiving upward private transfers tend to increase the amount of upward private transfers.

9. Both two correlation terms (corr(provide[id],amount[id], an estimate of correlation of random effects) was statistically significant (indicated in Table 6), which means that endogenous sample selection was an issue and should be adjusted. Specifically, the terms indicate that the unobserved individual-level factors that increase the probability of providing downward private transfers tend to decrease the amount of downward private transfers; while the unobserved observation-level (time-varying) factors that increase the probability of providing downward private transfers tend to increase the amount of downward private transfers.