Does Income Redistribution Reduce Inequality of Opportunities? Evidence from China

Abstract

This paper investigates whether and how income redistribution in China affects inequality of opportunity (IOp), defined as the share of income inequality attributable to circumstances beyond individual control. Using nationally representative data from the China Household Finance Survey (CHFS) and employing an ex-ante parametric approach with Shapley decomposition, we analyze the effects of three redistributive channels: taxation, government transfers, and inter-household transfers. The results show that taxation modestly reduces both inequality of outcome (IO) and IOp. In contrast, government transfers, particularly pensions, increase IOp due to institutional segmentation associated with the hukou system. Inter-household transfers also contribute to higher IOp by reinforcing intergenerational advantages. Additionally, we find that the classification of pensions significantly alters the redistribution’s measured impact. When pensions are treated as deferred income rather than government transfers, the second distribution reduces IOp more substantially. These findings suggest that redistributive policy effectiveness depends not only on the magnitude of redistribution but also on its institutional design and classification logic. The study provides new evidence on how fiscal and informal transfers affect structural inequality and calls for greater conceptual clarity in redistribution evaluation frameworks.

1. Introduction

The persistence of economic inequality has renewed scholarly and policy debates on the redistributive effectiveness (Cingano 2014; Doerrenberg and Peichl 2014; Atkinson 2015). Yet a fundamental question remains unresolved: Do policies that reduce income inequality also advance equality of opportunity? Different from the inequality of outcome (IO), recent literatures increasingly acknowledged that inequality of opportunity (IOp)—which reflects differences stemming from circumstances beyond an individual’s control, such as parental background, gender, or ethnicity—has become a central normative criterion for assessing distributive justice (Ferreira and Peragine 2015; Ramos and Van de Gaer 2016) and has been increasingly adopted as an empirical metric in economic research and policy analysis (Brunori et al. 2013; Golley et al. 2019; Brunori et al. 2023; Preuss et al. 2025), following Roemer’s (1998) seminal work.

While fiscal tools such as taxation and transfers are known to compress observed income gaps or IO (Atkinson 2003; Werning 2007; Caminada et al. 2012; Causa and Hermansen 2018; Lustig and Wang 2020), a critical gap persists in understanding whether these mechanisms address the more ethically troubling dimension of inequality (or IOp), with only a few notable exceptions (Roemer et al. 2003; Dunnzlaff et al. 2011). This analytical gap is particularly problematic for redistribution assessment because policies that successfully compress income gaps may simultaneously reinforce group-based disadvantages, if they fail to account for structural barriers embedded in institutional design (Corus et al. 2016; Powell 2008; 2024).

The stakes of this oversight are especially high in transitional economies like China, where rapid marketization coexists with persistent institutional segmentation through the household registration (hukou) system. This system creates a dual welfare structure that segments access to pensions and social benefits based on rural–urban residency (Zhang and Treiman 2013; Wu and Zheng 2018; Cai and Yue 2020). While China’s redistributive mechanisms have significantly reduced overall IO, with fiscal policies substantially lowering the Gini coefficient over the past decade (Lugo et al. 2024), their effect on IOp remains poorly understood. This institutional context raises critical concerns about redistributive effectiveness. When institutional segmentation shapes how redistributive benefits are allocated, China’s dual welfare structure may systematically exclude rural populations from redistributive gains, undermining the country’s equality of opportunity objectives despite impressive income compression achievements.

To address these concerns, this paper examines three key questions: (1) How do redistributive mechanisms affect inequality of opportunity in China? (2) What are the differential impacts of taxation, government transfers, and inter-household transfers on IOp? (3) Which circumstantial factors drive changes in IOp through redistribution?

We find that while redistributive mechanisms effectively reduced IO, they increased IOp by about 1.3 percentage points across all channels under the baseline evaluation framework that treats contributory pensions as government transfers. Decomposing this aggregate effect reveals important variation across redistribution channels. Specifically, taxation reduces both inequality measures (reduces IOp by 1.1 percentage points), while government transfers and inter-household transfers increase IOp through institutional barriers and inherited advantage reinforcement (+1.9 and +0.5 percentage points, respectively). Among government transfers, pensions emerge as the dominant driver of the overall regressive effect, accounting for more than the entire increase in IOp from government transfers. This regressive effect occurs because pensions function as institutional amplifiers rather than equalizers in China due to stratification from the pension-hukou system. This amplified parental hukou status contribution to IOp by 2.8 percentage points, while reducing the contribution of parental education by 1.7 percentage points. Critically, this pattern reveals that pensions operate more like deferred income—preserving and amplifying pre-existing labor market segmentation—rather than functioning as redistributive transfers that equalize opportunities across circumstantial groups. Given this institutional evidence, we propose an alternative evaluation framework that treats pensions as deferred income rather than government transfers. Under this reclassification, formal redistribution demonstrates substantial effectiveness in reducing IOp by nearly 3 percentage points. These findings highlight that the apparent counterproductive effect stems primarily from evaluation framework choices rather than fundamental redistributive failures, suggesting that proper conceptual frameworks are as critical as policy design in assessing redistributive effectiveness.

This paper makes three key contributions to the literature. First, it shifts the evaluative lens from inequality of outcome (IO) to inequality of opportunity (IOp) in redistributive research, addressing a fundamental gap where developing economy redistribution literature remains heavily skewed toward income compression metrics. By demonstrating how conventional redistribution evaluation frameworks may overlook structural disadvantages, this study advances a more nuanced understanding of redistributive effectiveness in transitional economies with institutional segmentation, offering critical insights for policy design where formal rules coexist with persistent stratification mechanisms.

Second, it expands the scope of redistributive analysis by incorporating informal mechanisms, the so-called third distribution, alongside conventional state-mediated instruments (such as taxes and government transfers), especially in opportunity-based frameworks. This broader accounting framework enables a more comprehensive empirical assessment of how both formal and informal redistributive channels shape opportunity structures. By documenting how private familial transfers often reinforce rather than mitigate inherited advantage, the paper brings to light a frequently overlooked mechanism that sustains structural inequality across generations, challenging the implicit assumption in redistribution literature that informal transfers serve as equity-enhancing supplements to formal welfare systems (Cox and Jimenez 1990; Rapoport and Docquier 2006; McKenzie and Rapoport 2007; Cai and Evans 2018; Boltz et al. 2019).

Finally, it contributes novel empirical evidence on how the classification of contributory pensions in redistributive accounting significantly alters opportunity-based assessments of redistribution, following the CEQ framework1 (Lustig 2018). The opposing redistributive effects on IOp, depending on pension classification, highlight that pension categorization is not merely a conceptual specification, but one that could profoundly impact the redistribution evaluation. This analysis contributes to the broader literature on IOp and fiscal redistribution by demonstrating the importance of classification sensitivity in evaluating equity through the lens of IOp.

The remainder of this paper is organized as follows. Section 2 provides institutional background on China’s redistributive framework and the hukou system. Section 3 describes our data from the China Household Finance Survey and explains our ex-ante parametric methodology with Shapley decomposition. Section 4 presents baseline results showing how different redistribution channels affect IOp. Section 5 examines the specific role of pensions and tests alternative classifications. Section 6 explores regional heterogeneity across eastern, central, and western China. Section 7 concludes with policy implications and suggestions for reform.

2. Background and Institutional Context

2.1. Conceptual Foundations of Inequality of Opportunity

The conceptual foundation of inequality of opportunity (IOp) originates in the normative distinction between circumstances and effort. Circumstances are factors beyond an individual’s control—such as parental background, place of birth, gender, or ethnicity—while effort refers to choices for which individuals can reasonably be held responsible, such as labor supply or savings decisions (Roemer 1998; Fleurbaey and Peragine 2013; Roemer and Trannoy 2016). This moral distinction—between “circumstances” and “effort”—has become central to debates in distributive justice, economics, and social policy. It underpins what is often referred to as the “level playing field” ideal: individuals should start from comparable positions, even if they end up with different outcomes due to effort. IOp reframes traditional equality debates by shifting focus from outcome-based equality to fairness in opportunity structures (Ramos and Van de Gaer 2016; Ferreira and Peragine 2015). The core ethical principle of this framework is that social arrangements should neutralize the effects of circumstances, ensuring that outcomes reflect effort alone. This aligns with the theory of luck egalitarianism, which holds that inequality is unjust when it stems from brute luck, but acceptable when it reflects option luck or responsible choice (Arneson 1989; Dworkin 1981). Accordingly, a central policy objective in promoting equity is to decouple economic outcomes from inherited or socially constructed disadvantages.

Empirically, IOp is often estimated using an ex-ante parametric approach that models income as a function of observed circumstances (Brunori et al. 2013; Golley et al. 2019), yielding the share of inequality explained by these factors (Ferreira and Gignoux 2011). This approach allows construction of counterfactual income distributions under full equality of opportunity, though estimates typically serve as lower bounds due to potential unobserved circumstantial factors (Hufe et al. 2022).

Beyond ethical considerations, IOp holds political significance. Survey and experimental studies find that citizens are more likely to support redistribution when inequality stems from circumstances rather than effort (Gaertner and Schokkaert 2012; Preuss et al. 2025). This insight underscores that addressing IOp is not only normatively desirable but also central to securing the political legitimacy of redistributive institutions.

2.2. Redistributive Architecture in China

China’s redistributive structure comprises three analytically distinct layers. The first distribution consists of income generated through market activity, including wages, capital returns, and business earnings. The second distribution is mediated through the fiscal system, comprising taxation and social transfers such as pensions, subsidies, and health reimbursements. The third distribution includes informal and voluntary inter-household transfers, such as family support, holiday gifts, educational aid, and community-based charity. The ‘first–second–third distribution’ terminology is not original. It reflects long-standing Chinese policy discourse and domestic scholars’ discussion. The ‘third distribution’ is often traced to Yining Li (1994)2 and has recently been re-emphasized in discussions of ‘common prosperity’ (Sun and Cao 2022).

Among these, the second distribution plays a central role in state-led redistribution. While earlier studies document that fiscal interventions have reduced income inequality (Lustig and Wang 2020; Lugo et al. 2024), the equalizing impact on IOp remains unclear. A central issue is the segmented nature of China’s welfare system, especially as it relates to the hukou registration system. Urban hukou holders have privileged access to pensions and public services, while rural and migrant populations face barriers in both coverage and benefit levels (Wu and Zheng 2018).

Public pensions are particularly emblematic of this segmentation. China’s system includes two main schemes: the Basic Pension Scheme for Urban Employees (BPSUE), a contributory program tied to formal employment, and the Basic Pension Scheme for Rural and Non-Working Urban Residents (BPSRNUR), which is partially subsidized and largely voluntary. Although pensions help reduce poverty among elderly households (Li et al. 2020), BPSUE disproportionately benefits higher-income, formally employed individuals (Cai and Yue 2020) and functions more like deferred income than progressive redistribution3 (Lustig 2018). In contrast, BPSRNUR’s limited coverage weakens its equalizing impact. These institutional asymmetries reinforce labor-market segmentation and mirror broader hukou-based disparities, thereby diminishing the redistributive effectiveness of the second distribution and raising critical concerns about its implications for inequality of opportunity. Regional disparities in coverage and benefit generosity further underscore the need for a more uniform and inclusive social security architecture (Fang and Feng 2020).

Moreover, this institutional complexity further reveals fundamental conceptual questions in policy evaluation frameworks (Lubell 2013; Domorenok et al. 2021)—specifically, how contributory social insurance systems should be categorized when assessing redistributive effectiveness (Hindriks and De Donder 2003; Bartels and Neumann 2021). The treatment of contributory pensions as government transfers versus deferred market income represents not only a technical classification choice, but also reflects deeper conceptual assumptions about the nature of redistribution itself (Inchauste and Lustig 2017; Lustig 2018; 2020), with potentially profound implications for policy evaluation conclusions about equality of opportunity.

In addition, redistribution extends beyond market earnings and formal government transfers. A third channel, known as the so-called third distribution, comprises income flows outside formal institutions and may play an important but understudied role in shaping inequality. In the Chinese context, this may include financial support from extended family members, informal assistance from neighbors or friends, and occasional charitable donations for education, healthcare, or hardship relief. These transfers are typically voluntary, unregulated, and non-contributory, reflecting relational, cultural, or moral obligations rather than legal entitlements. The redistributive implications of the third distribution are theoretically contested and empirically underexplored. On the one hand, wealthier households are better positioned to offer sustained material support across generations, thereby reproducing structural advantage and widening opportunity gaps (Corak 2013). On the other hand, philanthropic or community-driven transfers—although modest and institutionally underdeveloped in China—may serve as a supplementary channel of redistribution, particularly in areas underserved by state provision. Enjolras et al. (2018) provide a conceptual framework illustrating how such non-state initiatives, including charitable giving and social assistance, can contribute to distributive equity by addressing gaps left by formal welfare systems. Although empirical evidence remains limited4, the conceptual distinction between familial and philanthropic redistribution helps illuminate the heterogeneous functions that non-state transfers may play in shaping IO and IOp.

3. Materials and Methods

3.1. Income and Circumstance Variables

This paper utilizes data from the China Household Finance Survey (CHFS) 2013 (A full justification for the 2013 data wave and the institutional persistence underpinning our design is provided in Section 3.3). Unless otherwise noted, all income variables are annual amounts (RMB/year) referring to the preceding calendar year (2012) as per CHFS data documentation and survey reports. The main income variables include pre-tax wage income5, post-tax wage income, inter-household transfer income, public transfer income, and total income after redistribution, as shown in

Table 1. Descriptive statistics for main variables (CHFS 2013).

	Mean	Std. Dev
Income Variables (Yuan)
1st Distribution	18,864.26	28,649.56
2nd Distribution	453.40	7807.83
#Tax and Fees	−1065.04	3762.77
#Public Transfer	1518.44	6897.4
#Government Subsidies	73.69	717.64
#Agricultural Subsidies	114.01	5564.9
#Pensions	904.78	3314.84
#Unemployment Relief	29.27	767.26
#Healthcare Reimbursement	396.69	1607.8
3rd Distribution	772.92	4129.45
Total Income After Redistribution	20,090.59	28,206.94
Circumstance Variables
Gender (male = 1)	0.572	0.495
Ethnics (Han = 1)	0.936	0.245
Father’s
Education level	6.887	4.571
CPC Membership	0.291	0.454
Hukou type (urban = 1)	0.485	0.500
Administration level (fuke = 1)	0.053	0.224
Mother’s
Education level	5.155	4.550
CPC Membership	0.066	0.249
Hukou type (urban = 1)	0.424	0.494
Administration level (fuke = 1)	0.008	0.087
Observations	8772	—

Notes: (1) Income variables are annual (RMB/year); (2) for CHFS 2013, refer to calendar year 2012; (3) “Education level” measures total completed years of formal schooling from primary onward.

. Among them, inter-household transfer income refers to various forms of financial support from non-family members6. This includes holiday-related income, income from life events, financial assistance for education or medical expenses, living allowances7, inheritance, charitable contributions8, and informal financial assistance from friends or community members. Subsidies to disadvantage groups (government subsidies) is one of the main source in public transfer income, including subsidies for impoverished households9, allowances for households under the Five Guarantees Program10, financial compensation or pensions for bereaved families of deceased workers or military personnel, relief funds for poverty alleviation, disaster relief payments for natural or man-made catastrophes, food subsidies, financial incentives during the implementation of the one-child policy (e.g., housing subsidies or direct payments), subsidies for reforestation efforts under national ecological restoration programs, minimum living standard allowances (a means-tested welfare program providing basic income support in China), educational subsidies (e.g., scholarships, tuition waivers, or financial aid for rural students), and housing assistance (e.g., rental subsidies or public housing support for low-income families). Other Public transfer income includes agricultural subsidies, pensions11, unemployment insurance (unemployment benefits), and medical insurance reimbursements.

Although inter-household transfer income may not fully capture the third distribution due to possible omissions, it serves as a reasonable proxy, as it reflects income transfers occurring outside market mechanisms and government redistribution. Accordingly, disposable income can be categorized into three sectors: the first distribution12 (pre-tax wage or market income), the second distribution (taxes and government transfers), and the third distribution (inter-household transfer income), as shown in Table 1. While most IOp studies use individual earnings as the target variable, we use the household as the accounting unit and household income per capita (HIPC)13 as the target variable for IOp. This choice follows Yang et al. (2021)14 and fits the CHFS reporting structure, where many redistributive inflows (e.g., subsidies and inter-household transfer) are recorded at the household level and are most coherently analyzed on a per-capita basis.

We acknowledge that using HIPC departs from the canonical IOp practice that targets individual earnings; however, it is the appropriate construct here because redistribution operates through household-reported items and because intra-household pooling and marriage sorting transmit individual circumstances into household income arrangements in China.

In line with the IOp literature (Bourguignon et al. 2007; Golley et al. 2019), we restrict our analytical sample to working-age individuals (26–60 years old) who are household heads and have complete information on key circumstance variables (e.g., parental hukou, parental education, and other family background characteristics). This yields 8772 individuals out of more than 89,000 observations in the full survey sample. Such restrictions are standard in IOp studies because they help exclude outliers from the non-working population and ensure that income measures are appropriately matched with relevant circumstance variables. Noting that although we divergent from traditional IOp literature by accounting at household level and use HIPC as target variable, the sampling restriction rule still applies for two reasons: (i) it creates a clear mapping between the measured circumstance vector (defined for the head, e.g., parental education and hukou) and the household outcome we study; and (ii) in CHFS the household head reports the most complete information on key background variables, yielding higher data completeness and comparability across households. Note that pensions and other income received by any co-resident member are aggregated to the household total and, therefore, enter HIPC; hence, retirees’ pensions are not excluded from our outcome measure with the sample restriction. We refer to Section 3.2 for the co-residence and income-pooling rationale. We acknowledge that using the head’s parental background as a proxy for household circumstances departs from the ideal of member-specific circumstances, but it is the most consistent and data-complete approach available for linking circumstances to HIPC in CHFS.

Moreover, this approach inevitably reduces representativeness, as the selected subsample may differ from the full sample in its distribution of certain circumstances. For instance, the parental urban hukou proportion is lower in the analytical sample than in the full sample, where the proportions are 29.3% for fathers and 24.8% for mothers. This under-representation, however, stems from inherent data requirements of the ex-ante parametric approach used for IOp estimation, which necessitates complete circumstance information for all included observations. To address possible sensitivity to the parental hukou composition of the analytic sample, we conducted two compact checks. First, we reweighted the data to ensure that the joint parental hukou distribution (RR, RU, UR, UU) is balanced. Second, we re-estimated the models on homogeneous subsamples (RR-only; UU-only). Full results are reported in Appendix A.1 

Table A1. Shapley’s share of parental hukou.

Income	Baseline (swgt)	EqualWeight-Hukou	Δ(EW-Base)	% Change	RR-Only	UU-Only
1st Dis	0.0553	0.0394	−0.0159	−28.8%	0.0000	0.0071
2nd Dis	0.0829	0.0566	−0.0263	−31.7%	0.0000	0.0072
#Tax	0.0505	0.0361	−0.0144	−28.5%	0.0000	0.0072
#PT	0.0875	0.0579	−0.0269	−33.8%	0.0000	0.0074
3rd Dis	0.0569	0.0419	−0.0150	−26.4%	0.0000	0.0069
Total	0.0840	0.0544	−0.0296	−35.2%	0.0000	0.0072

Note: Entries are Shapley shares for the parental-hukou circumstance set (father’s and mother’s hukou jointly). Baseline (swgt) uses survey sampling weights. EqualWeight–hukou balances the four joint parental-hukou cells (RR, RU, UR, UU) at 25% each. RR-only and UU-only restrict to households where both parents share the same hukou type; the small positive values in UU-only are negligible relative to the baseline. Levels fall under equal weighting and are (near) zero in homogeneous subsamples.

.

Choosing circumstance variables is critical for estimating IOps (Bourguignon et al. 2007; Marrero and Rodríguez 2013). Following recent studies (Golley et al. 2019; Yang et al. 2021; Dai and Li 2021), we select gender (male = 1), ethnicity (Han = 1), and indicators of family background as key circumstance variables. The family background variables include parental education level, hukou type (urban = 1), Communist Party membership, and the highest administrative rank attained by parents (with Deputy-Section Head level or above coded as fuke = 1). These variables are used to estimate IOp levels before and after redistribution.

It is important to note that three commonly used circumstance variables (age, individual hukou status, and region of residency) are excluded from our analysis. First, although age is undoubtedly beyond an individual’s control, its inclusion as a determinant of IOp remains subject to debate15. One of the issues lies in disentangling the effects of “age” from “work experience”, which is a key explanatory variable in the Mincer earnings equation (Mincer 1974). Given the significant influence of work experience on income, further examination is required to determine whether age should be treated as an exogenous variable contributing to IOp (Ramos and Van de Gaer 2016; Brunori 2017). Second, an individual’s household registration status (hukou) is not entirely beyond their control. Individuals can convert from rural to urban hukou through various channels16, including employment, marriage, education, and economic or professional achievement (Zhang and Treiman 2013). Since most of these could be achieved through personal effort, we argue that an individual’s hukou status should not be classified as a circumstance variable beyond one’s control. Third, although the region of birth is often regarded as a circumstantial factor beyond an individual’s control (Golley et al. 2019), using current residence as a proxy could be problematic, as migration decisions could be made by the individual. Prior studies classify migration as an effort variable, as it is at least partially within an individual’s control (Bourguignon et al. 2007; Hufe et al. 2022). In the absence of data on the region of birth in the CHFS dataset, we argue that current residence is more appropriate for heterogeneity analysis. This argument aligns with the approach of Marrero and Rodríguez (2013), who used data from 50 U.S. states to explore the relationship between inequality of opportunity and economic growth.

Some other potentially relevant circumstances (e.g., region/place of birth and parental/childhood wealth) are not included due to data limitations and conceptual concerns. We do not observe a reliable birthplace for the full analytic sample; current residence or current hukou locality may reflect post-childhood migration and thus mix circumstances with outcomes. Likewise, contemporaneous household wealth can reflect the respondent’s own savings and returns and is therefore an intermediate outcome, not a pure exogenous circumstance. In line with ex-ante IOp measurement, we avoid post-determination channels.

In addition, we treat parental hukou and parental education as predetermined circumstances in the ex-ante opportunity framework. We acknowledge that these variables can be correlated with unobserved family endowments—such as inherited social networks, information, or informal institutional access—that also affect adult incomes. Our Shapley decomposition (which will be introduced in Section 3.4), therefore, provides a descriptive accounting of inequality attributable to the observed circumstance set; it does not identify causal effects of any single circumstance holding all else constant. To guard against attribution bias arising from unobserved family endowments, we conduct a robustness analysis that expands the circumstance set to include parental occupation categories and indicators for whether either parent served as a village cadre (i.e., “CunGanBu”), as a proxy for inherited social capital. Results from this expanded specification are reported in Section 4.2 and Section 4.4 and Appendix A.1 

Table A2. Total inequality and IOp before and after redistribution (with additional circumstance variables).

	1st Dis	2nd Dis	#Tax	#Public	3rd Dis	Total
IO
MLD	0.5259	0.4781	0.5138	0.5020	0.5156	0.4706
Gini	0.5265	0.5052	0.5183	0.5131	0.5242	0.5264
IOp
Absolute       (IOA)	0.0957 (0.005)	0.0917 (0.005)	0.0879 (0.005)	0.0993 (0.005)	0.0968 (0.005)	0.0929 (0.005)
Relative       (IOR)	0.1810	0.1908	0.1702	0.2013	0.1867	0.1965

Note: (1) Bootstrap standard errors are in parentheses; (2) Gini is reported for comparability; IOp metrics (IOA and IOR) and all Shapley decompositions are computed using MLD. Data Source: CHFS 2013 and authors’ calculation.

and

Table A3. Shapley decomposition before and after redistribution (with additional circumstance variables).

	1st Dis	2nd Dis	#Tax	#Public	3rd Dis	Total
Gender	0.0062 (3.41%)	0.0056 (2.94%)	0.0065 (3.89%)	0.0052 (2.61%)	0.0058 (3.08%)	0.0051 (2.61%)
Ethnicity	0.0030 (1.66%)	0.0027 (1.40%)	0.0028 (1.62%)	0.0029 (1.42%)	0.0030 (1.63%)	0.0027 (1.37%)
Parental
Education	0.0733 (40.58%)	0.0531 (27.83%)	0.0692 (40.74%)	0.0572 (28.44%)	0.0760 (40.77%)	0.0561 (28.57%)
CPC	0.0055 (3.03%)	0.0057 (2.96%)	0.0051 (3.03%)	0.0059 (2.93%)	0.0056 (2.98%)	0.0058 (2.94%)
Hukou	0.0286 (15.82%)	0.0415 (21.78%)	0.0261 (15.35%)	0.0440 (21.87%)	0.0290 (15.55%)	0.0417 (21.24%)
Admin level	0.0076 (4.21%)	0.0085 (4.47%)	0.0074 (4.37%)	0.0087 (4.34%)	0.0074 (3.94%)	0.0083 (4.23%)
Occupation
Farmer	0.0254 (14.09%)	0.0340 (17.81%)	0.0236 (13.91%)	0.0357 (17.75%)	0.0268 (14.38%)	0.0352 (17.94%)
Employed	0.0245 (13.59%)	0.0325 (17.05%)	0.0228 (13.44%)	0.0340 (16.92%)	0.0258 (13.83%)	0.0337 (17.16%)
Self-Emp	0.0033 (1.82%)	0.0025 (1.30%)	0.0032 (1.91%)	0.0026 (1.28%)	0.0038 (2.04%)	0.0030 (1.52%)
Chore	0.0023 (1.25%)	0.0033 (1.72%)	0.0020 (1.20%)	0.0035 (1.76%)	0.0023 (1.26%)	0.0033 (1.68%)
CunGanBu	0.0009 (0.53%)	0.0014 (0.73%)	0.0009 (0.54%)	0.0014 (0.70%)	0.0010 (0.54%)	0.0014 (0.73%)
IOR	0.1810 (100%)	0.1908 (100%)	0.1702 (100%)	0.2013 (100%)	0.1867 (100%)	0.1965 (100%)

Note: Relative shares are included in parentheses.

.

3.2. Parental Pension as Conditional Redistribution: A Functional Perspective

In analyzing household income composition in China, particularly in studies of the redistributive effect, parental pension income occupies an ambiguous yet analytically important position. On one hand, pensions are disbursed through state-administered social insurance schemes, often funded by a combination of individual contributions and government subsidies (Fang and Feng 2020). From this institutional perspective, they may be interpreted as part of the broader system of secondary distribution17, especially when they substitute for direct transfers or other state-provided benefits that are absent for younger adults in the same household (Cai and Yue 2020; Lugo et al. 2024).

This interpretation gains further plausibility in contexts where multigenerational co-residence leads to significant intra-household income pooling, a widespread norm in China (Park et al. 2012). In such settings, elderly pensions may serve not merely as income for the recipient but as household-level fiscal support that indirectly channels public resources toward younger, potentially disadvantaged members (Shan and Park 2023; Guo et al. 2025). For instance, in families where adult children lack access to stable employment or public transfers, parental pensions may functionally compensate for this institutional gap. Under this configuration, they can be viewed as an indirect, state-mediated redistribution channel.

However, this view is not definitive. An alternative interpretation—one that we explore elsewhere in this study—is to treat pension income as deferred earnings, reflecting earlier labor market contributions rather than current redistributive intent (Lustig 2018). From this angle, pensions are more akin to accumulated private claims on prior productivity, and their inclusion as redistribution income becomes theoretically contested. The dual character of pensions—institutionally public yet rooted in individual market contributions—necessitates a cautious and context-specific classification.

Therefore, in this analysis, we adopt a functional approach: parental pensions are tentatively included in the redistribution income category insofar as they play a compensatory role within the household, particularly for non-earners or underprivileged adult members. This classification does not preclude alternative interpretations and will be revisited in later sections to account for heterogeneity in pension origins, entitlements, and usage.

3.3. Data Selection and Institutional Persistence Justification

Although the 2013 CHFS may appear dated, it remains the most methodologically appropriate dataset for this study, for two reasons:

First, it provides the most sufficient sample size in the later waves. CHFS 2013 provides comprehensive parental information for 8772 households, including detailed circumstance variables (parental education, hukou status, CPC membership, and administrative rank) necessary for accurate IOp estimation. While later CHFS waves (2015, 2017, 2019) continued to collect parental background information, most of these data are null, yielding insufficient sample sizes (<3000 observations) for robust estimation.

Second, it provides a more detailed disaggregation of income sources. Our research design requires a detailed breakdown of redistributive channels, which are unavailable in alternative Chinese household surveys. CHFS 2013 uniquely provides the granular categorization essential for our analysis, as mentioned in Section 3.1. Alternative datasets such as China Household Income Project (CHIP) 2018, Chinese General Social Survey (CGSS) 2017–2021, and China Family Panel Studies (CFPS) 2016–2022 provide more recent observations but lack this level of redistributive detail necessary for decomposing second and third distribution effects on IOp.

While using 2013 data limits temporal generalizability, the core institutional mechanisms underlying hukou-pension segmentation between urban and rural populations have remained structurally persistent to the current date, suggesting our findings retain policy relevance. Despite incremental policy reforms under China’s urbanization strategy (2014–2020) aimed at addressing welfare disparities (Sieg et al. 2020), the hukou system continues to perpetuate systematic disadvantages for rural populations in accessing urban social benefits and pension security (Butchko 2024). Recent analyses confirm that institutional arrangements still generate significant disparities in social protection between urban and rural hukou holders (Wu and Zhang 2018; Hung 2022), with the dual-track pension system maintaining fundamental segmentation structures that disadvantage rural residents (Yang and Zhao 2024). These ongoing institutional barriers demonstrate that the fundamental segmentation structure identified in our analysis remains relevant for contemporary policy discussions.

3.4. Methodology

We follow the ex-ante parametric method proposed by Ferreira and Gignoux (2011). Individual income $Y$ is determined by circumstance variables $C$ (beyond individual control), effort variables $E$ (within individual control), and random factors $u$, such as luck:

$$Y_i=f(C_i,E_i,u_i)$$

(1)

Effort is often shaped by circumstance factors. For instance, children from wealthier backgrounds tend to have greater access to quality education, which can influence their efforts and outcomes. Thus, circumstantial factors can directly or indirectly shape personal effort, which subsequently affects income (Bourguignon et al. 2007; Ferreira and Gignoux 2011). Equation reflects these interdependencies can be rewritten as follows:

$$Y_i=f\left(C_i,g\left(C_i,v_i\right),u_i\right)$$

(2)

The structural equation is as follows:

$$Y_i=\alpha +\beta C_i+\gamma E_i+u_i$$

(3)

The relationships between income and environmental variables, and income and effort variables, are measured by coefficients $\beta$ and $\gamma$. Assuming effort is endogenous to the circumstance and unobservable factors as follows:

$$E_i=\delta C_i+{\eta }_i$$

(4)

By substituting Equation (4) into Equation (3), we get the following:

$$Y_i=\alpha +\left(\beta +\gamma \delta \right)C_i+\gamma {\eta }_i+u_i$$

(5)

This can be simplified as follows:

$$Y_i=\alpha +\theta C_i+{\epsilon }_i,with \theta \equiv \beta +\gamma \pi ,{\epsilon }_i\equiv \gamma {\eta }_i+u_i$$

(6)

Consequently, if any relevant circumstance is unobserved, its effect is absorbed into the residual together with effort, implying that our IOp measures should be interpreted as lower bounds (see also Section 2.1).

In empirical terms, Equations (1)–(6) implement the standard ex-ante projection of income on observed circumstances (Ferreira and Gignoux 2011). This can be viewed as the reduced-form counterpart of a Mincer-type earning model, where own schooling and experience are effort-proxied channels and therefore intentionally not included among regressors in the ex-ante IOp setup. The estimated coefficients on circumstances capture their total association with income (direct and effort-mediated), while the residual comprises both effort and any unobserved circumstances.

Equation (6) can be estimated using Ordinary Least Squares (OLS), where $\rho$ represents the total effect of circumstance variables on income (including both direct and indirect effects mediated through effort), and ${\omega }_i$ is the sum of unobserved random variables and error terms. Using the estimated coefficients, a counterfactual income distribution18 under complete equality of opportunity can be constructed:

$$\widehat{Y_i^{\mathrm{EO}}}=\widehat{\mathsf{\alpha }}+\widehat{\mathsf{\theta }}{c}_i$$

(7)

Under the assumption of complete equality of opportunity, income differences among individuals should not be influenced by circumstantial differences. The income fitted for individuals grouped by the same circumstance should be uniform, enabling the measurement of absolute and relative inequality of opportunity (IOA and IOR, respectively):

$$IOA=I\left(\left\{{\widehat{Y}}_i\right\}\right)$$

(8)

$$IOR={\displaystyle \frac{I\left(\left\{{\widehat{Y}}_i\right\}\right)}{I\left(\right\{{\widehat{Y}}_i^k\left\}\right)}}$$

(9)

where $I$ is measured using the generalized entropy index GE(0), also known as the Mean Log Deviation (MLD) index. For a sample of strictly positive incomes $Y_i>0$ with mean $\mu$, the GE index is as follows:

$$GE(\mathsf{\kappa })=\left\{\begin{array}{ll}{\displaystyle \frac{1}{n\mathsf{\kappa }(\mathsf{\kappa }-1)}}{\displaystyle \sum _{i=1}^n}\left[{\left({\displaystyle \frac{y_i}{\mu }}\right)}^{\mathsf{\kappa }}-1\right],& \mathsf{\kappa }\ne 0,1,\\ {\displaystyle \frac{1}{n}}{\displaystyle \sum _{i=1}^n}\mathrm{l}\mathrm{n}\left({\displaystyle \frac{\mu }{y_i}}\right),& \mathsf{\kappa }=0 \left(\mathrm{MLD}\right),\\ {\displaystyle \frac{1}{n}}{\displaystyle \sum _{i=1}^n}\left({\displaystyle \frac{y_i}{\mu }}\right)\mathrm{l}\mathrm{n}\left({\displaystyle \frac{y_i}{\mu }}\right),& \mathsf{\kappa }=1 (\mathrm{Theil}T).\end{array}\right.$$

(10)

We adopt GE(0) as our baseline for IOp because (i) it is additively decomposable, which is essential for ex-ante IOp measurement where the counterfactual fitted distribution is constructed from circumstances (Ferreira and Gignoux 2011); (ii) it is more sensitive to disparities in the lower tail, aligning with our focus on opportunity disadvantages; (iii) it is widely used in the IOp literature, facilitating comparability (e.g., Ferreira and Gignoux 2011; Ramos and Van de Gaer 2016; Golley et al. 2019).

Furthermore, we employ Shapley decomposition to measure the contribution of various circumstance variables to IOp (Shorrocks 2013). Let $N$ denote the set of circumstance variables and $I\left(S\right)$ the IOp (using GE(0)) computed with the subset $S\subseteq N$ included in the ex-ante model. The Shapley value for circumstance $j\in N$ is as follows:

$${\varphi }_j=\sum _{S\subseteq N\backslash \left\{j\right\}}{\displaystyle \frac{\left|S\right|!\hspace{0.17em}\left(\right|N|-|S|-1)!}{\left|N\right|!}}\hspace{0.17em}\left[I(S\cup \{j\left\}\right)-I\left(S\right)\right].$$

(11)

This is the average marginal contribution of $j$ across all orderings. We implement this by averaging across many random permutations; the contributions sum exactly to the total IOp.

This method is particularly suitable for analyzing IOp for two main reasons: (1) it fully accounts for the potential interactions and ordering of circumstance variables, ensuring a comprehensive decomposition; and (2) it guarantees that the sum of all circumstantial factor contributions equals the total IOp (Ramos and Van de Gaer 2016). Shapley decomposition calculates the average marginal contribution of each circumstance variable to IOp by considering all possible combinations and sequences in which the variables can be introduced. Although the results reflect correlations rather than causal relationships, the method remains robust to statistical challenges such as multicollinearity among circumstance variables. As a result, Shapley decomposition offers a nuanced and systematic understanding of the relative importance of each circumstance variable in shaping inequality of opportunity (Shorrocks 2013).

Having estimated absolute and relative IOp for each income concept (market income, income after taxes/transfers, and total disposable income), we next introduce the effect of redistribution on inequality using a redistributive-effect (MT-type) difference. Following the classic Musgrave–Thin (MT) index (Musgrave and Thin 1948), the following can be established:

$$M{T}_{IO}=I{O}^{pre}-I{O}^{post}=G{E}_0\left(Y^{pre}\right)-G{E}_0\left(Y^{post}\right)$$

(12)

where $I{O}^{pre}$ represents IO in market income, and $I{O}^{post}$ represents IO after redistribution. A positive MT indicates that the redistribution mechanism reduces overall inequality, while a negative MT suggests otherwise. Analogously for IOp, we evaluate the redistributive effect on IOp as follows:

$$M{T}_{IOR}=IO{R}^{pre}-IO{R}^{post}=[{\displaystyle \frac{G{E}_0\left(\widehat{Y}ᴱᴼ,pre\right)}{G{E}_0\left(Y^{pre}\right)}}]-[{\displaystyle \frac{G{E}_0\left(\widehat{Y}ᴱᴼ,post\right)}{G{E}_0\left(Y^{post}\right)}}]$$

(13)

where $IO{R}^{pre}$ represents the relative IOp (or the share of inequality attributable to circumstances) before redistribution, $IO{R}^{post}$ represents the relative IOp after redistribution, and $M{T}_{IOR}$ represents the impact of the redistribution mechanism on IOp.

4. Baseline Results

This section presents the core empirical findings regarding the relationship between income redistribution and inequality of opportunity (IOp) in China. We begin by assessing how individual circumstance variables are associated with income before and after redistribution. We then quantify how different redistributive mechanisms influence both inequality of outcome (IO) and IOp, and we examine which circumstances matter most in explaining unfair inequality.

4.1. Baseline Income Determinants

Table 2. Circumstantial determinants before and after redistribution.

	1st Dis	2nd Dis	#Tax	#Public	3rd Dis	Total
Gender	0.176 ***	0.160 ***	0.180 ***	0.157 ***	0.169 ***	0.152 ***
Ethnicity	0.144 ***	0.131 ***	0.133 ***	0.139 ***	0.141 ***	0.126 ***
Father’s:
Education years	0.029 ***	0.023 ***	0.028 ***	0.025 ***	0.030 ***	0.025 ***
CPC membership	0.052 **	0.044 *	0.049 *	0.044 *	0.050 *	0.042 *
Hukou	0.124 ***	0.165 ***	0.119 ***	0.170 ***	0.135 ***	0.172 ***
Administration level	0.010 *	0.127 **	0.104 *	0.124 **	0.083 *	0.113 **
Mother’s:
Education years	0.042 ***	0.030 ***	0.040 ***	0.032 ***	0.042 ***	0.031 ***
CPC membership	−0.098 *	−0.072	−0.091 *	−0.078	−0.093 *	−0.069
Hukou	0.258 **	0.353 ***	0.239 ***	0.368 ***	0.248 ***	0.342 ***
Administration level	0.307 **	0.255 *	0.310 **	0.253 *	0.308 **	0.258 *
Adj. R-squared	0.155	0.164	0.140	0.177	0.163	0.169

Note: * $p$ < 0.1, ** $p$ < 0.05, *** $p$ < 0.01. Data Source: CHFS 2013 and authors’ calculation.

reports OLS estimates of the association between circumstance variables and income across different redistribution layers. Column (1) presents results based on market income (i.e., pre-tax, pre-transfer earnings). Column (2) incorporates the second distribution as a whole, followed by separate specifications that isolate the effects of income taxes (Column 3) and government transfers (Column 4). Column (5) presents the inter-household transfers to capture the third distribution, and Column (6) presents results for total disposable income, incorporating all three layers of redistribution.

Across all specifications, gender, ethnicity, and parental education exhibit consistently significant positive associations with income. Specifically, male, Han-ethnic individuals with better-educated parents tend to have higher disposable income. These patterns reflect persistent structural advantages in labor markets and human capital acquisition. In contrast, parental Party membership and administrative rank show weaker or statistically insignificant effects, suggesting limited direct income advantage from political or bureaucratic lineage in the contemporary context.

Most notably, parental hukou status exerts a strong and increasing influence as we move from market income to disposable income. This pattern foreshadows the key institutional insight of this study: the redistributive structure may reinforce hukou-based advantages rather than mitigate them.

4.2. Aggregate Impact of Redistribution on IOp

Table 3. Total inequality and IOp before and after redistribution.

	1st Dis	2nd Dis	#Tax	#Public	3rd Dis	Total
IO
MLD	0.5259	0.4781	0.5138	0.5020	0.5156	0.4706
Gini	0.5265	0.5052	0.5183	0.5131	0.5242	0.5264
IOp
Absolute       (IOA)	0.0924 (0.006)	0.0881 (0.005)	0.0847 (0.005)	0.0956 (0.005)	0.0932 (0.006)	0.0889 (0.005)
Relative       (IOR)	0.1758	0.1842	0.1649	0.1947	0.1808	0.1889

Notes: (1) Bootstrap standard errors are in parentheses; (2) Gini is reported for comparability; IOp metrics (IOA and IOR) are computed using MLD. Data Source: CHFS 2013 and authors’ calculation.

presents our baseline estimates of absolute IOp (IOA) and relative IOp (IOR) across redistribution stages. Three findings stand out.

First, as expected, total income inequality declines steadily as redistributive layers are applied. The MLD drops from 0.526 in the market income stage to 0.471 after all redistributive means. This confirms that China’s redistributive system compresses overall income disparities, consistent with previous findings on fiscal incidence (Lustig and Wang 2020). It is worth noting that, to aid comparability with the broader inequality literature, we also report the Gini coefficient where relevant (e.g., Table 3). Our IOp estimation and Shapley attributions rely on GE(0) because it is additively decomposable across types and more sensitive to lower-tail changes. Accordingly, when Gini and GE(0) diverge, the IOp reading follows GE(0); for example, the near-flat Gini from market to total income is consistent with redistribution compressing the lower tail, which GE(0) registers more strongly.

Second, absolute inequality of opportunity (IOA)—which captures the variance in income attributable to exogenous circumstances—also falls, from 0.092 to 0.089. However, the decline is modest compared to the overall reduction in IO. This limited reduction suggests that the redistributive system only partially addresses the structural disadvantages embedded in familial background, hukou status, and demographic traits. It also reflects the fact that redistribution narrows some income gaps more effectively than others, and may bypass or inadequately compensate for deeply institutionalized sources of inequality.

Third, and most notably, relative inequality of opportunity (IOR)—presenting the share of total inequality attributable to circumstances—increases by 1.31 percentage points. While IOA falls slightly, it declines less rapidly than total income inequality, meaning that circumstances account for a larger proportion of the remaining inequality. This dynamic indicates a regressive shift in the composition of inequality: as redistribution reduces overall dispersion, the portion attributable to inherited or ascriptive factors becomes more dominant.

These findings reveal a central paradox: redistribution compresses income but makes unfair inequality more salient. The reason lies in the asymmetry of gains. Opportunity-advantaged groups—typically those with urban hukou and more educated parents—capture a disproportionate share of the redistributive benefits. Meanwhile, opportunity-disadvantaged individuals—whose constraints are structural and beyond personal control—see limited improvement. As a result, redistribution may inadvertently magnify the relative role of circumstance in explaining income differences.

To guard the sensitivity of these baseline results, we re-estimate IOp using an expanded set of circumstance variables that augments the baseline list (gender; ethnicity; parental education; parental hukou; parental CPC membership; parental administrative rank) with parental occupations (farmer, employed, self-employed, domestic/housework) and a proxy indicator for inherited social capital (Appendix A.1 Table A2). Results show that both absolute IOp (IOA) and relative IOp (IOR) increase modestly at all layers (e.g., Total IOR: 0.1889→0.1965, IOA: 0.0889→0.0929). The ranking across redistribution stages is unchanged: IOR is lowest after taxes and highest after public transfers (baseline: 0.1649 vs. 0.1947; expanded: 0.1702 vs. 0.2013), consistent with the main text. These shifts are small in magnitude and leave our interpretation intact.

4.3. Contrasting Taxes and Transfers

To uncover what drives this paradox, we disaggregate redistribution into second distribution components: income tax and government transfer.

Taxation exerts a modest but consistent equalizing effect. The application of personal income tax reduces both IO and IOp: the IOR declines by 1.1 percentage points relative to the market income stage. While China’s income tax system is characterized by relatively low marginal rates and limited coverage, particularly in rural areas and among informal-sector workers, the tax structure is nonetheless progressive in form (Benedek et al. 2022; Lugo et al. 2024). The limited but positive redistributive role of the tax structure suggests that even partial fiscal extraction from upper-income brackets can reduce unfair disparities in disposable income—especially when these disparities are strongly correlated with circumstance variables such as parental education and hukou.

In contrast, government transfers increase IOR by 1.89 percentage points, even as they reduce overall income inequality. This suggests that transfers disproportionately benefit individuals with more advantaged backgrounds, thereby increasing the relative contribution of circumstance to income inequality. Put differently, public transfers in China improve headline income metrics but simultaneously worsen the fairness of income generation and redistribution, at least when evaluated through the lens of IOp.

The regressive effect of transfers on IOp thus arises not from their intent but from their eligibility rules and benefit structures. In this context, redistribution operates as a reinforcement mechanism, disproportionately channeling resources to those already well-positioned in the opportunity hierarchy. Government transfer programs that are structurally tied to employment history and the urban hukou channel greater benefits to already advantaged groups, rather than correcting structural disadvantages.

These findings illustrate that evaluating redistributive instruments solely by their impact on outcome inequality can obscure important dimensions of fairness. From an IOp perspective, not all redistribution is created equal: its effectiveness depends not only on the amount redistributed but also on who benefits, through what mechanisms, and under what structural constraints. This observation sets the stage for the later section, which interrogates in greater detail the institutional role of China’s pension system in shaping redistributive outcomes.

4.4. Role of Third Distribution19

The final component of redistribution analyzed in this study is the third distribution, which encompasses informal inter-household transfers such as financial support from family members, holiday gifts, educational assistance, and other forms of non-market, non-state income. While often overlooked in fiscal incidence studies, these transfers are prevalent in China and represent a significant source of income supplementation, as shown in Table 1.

Table 3 illustrates that the IOR increases by 0.5 percentage points after these private transfers are included. Though smaller in magnitude than the impact of government transfers, the direction of the effect is consistent: inter-household transfers contribute to the amplification of opportunity inequality. This finding points to an important mechanism of private reproduction of structural advantage. Wealthier households—particularly those with urban hukou, stable formal employment, and higher educational capital—are better positioned to provide sustained financial support to younger generations. Such support may include tuition payments (Dong et al. 2019), housing down payments (Liao and Zhang 2021), startup capital (Lyu et al. 2022), or emergency assistance (Xu 2024). These transfers, while ostensibly “private,” function as extensions of familial privilege and compound the cumulative effects of advantageous circumstances. Their redistributive effect, therefore, is not neutral: rather than closing opportunity gaps, they tend to entrench them.

From a policy perspective, the third distribution occupies an ambiguous space: it lies outside the regulatory reach of the state, yet it significantly affects distributive outcomes (Cox et al. 2004; Doyle 2015). While some scholars have argued that informal transfers compensate for gaps in formal welfare provision (Cox and Jimenez 1990; Rapoport and Docquier 2006; McKenzie and Rapoport 2007; Cai and Evans 2018; Boltz et al. 2019), the evidence presented here suggests that such compensation is uneven at best and regressive at worst. In the absence of progressive public redistribution, the unequal capacity of households to provide support perpetuates opportunity gaps across generations. Moreover, the regressive impact of informal transfers on IOp underscores the limits of redistribution that relies on familial or community networks in structurally unequal societies. Where opportunity structures are already segmented by hukou, education, and parental occupation, informal redistribution may simply reproduce the social order rather than transform it. Redistribution occurring outside formal institutions is thus no substitute for equity-enhancing public policy.

4.5. Which Circumstances Matter Most?

To describe how background structure relates to measured IOp, we decompose the relative contributions of individual circumstance variables to IOR using the Shapley value method. This approach allows for a fair allocation of explanatory power across correlated variables by averaging their marginal effects over all possible orderings (Shorrocks 2013; Ramos and Van de Gaer 2016).

Table 4. Shapley decomposition before and after redistribution.

	1st Dis	2nd Dis	#Tax	#Public	3rd Dis	Total
Gender	0.0066 (3.74%)	0.0058 (3.17%)	0.0070 (4.26%)	0.0055 (2.81%)	0.0061 (3.39%)	0.0053 (2.83%)
Ethnicity	0.0014 (0.78%)	0.0013 (0.71%)	0.0012 (0.73%)	0.0014 (0.73%)	0.0013 (0.74%)	0.0013 (0.67%)
Parental
Education	0.0946 (53.83%)	0.0739 (40.13%)	0.0892 (54.07%)	0.0793 (40.75%)	0.0987 (54.60%)	0.0782 (41.42%)
CPC	0.0083 (4.74%)	0.0089 (4.85%)	0.0078 (4.75%)	0.0093 (4.78%)	0.0086 (4.73%)	0.0092 (4.85%)
Hukou	0.0549 (31.24%)	0.0824 (44.74%)	0.0500 (30.34)	0.0871 (44.71%)	0.0563 (31.15%)	0.0833 (44.12%)
Admin level	0.0100 (5.67%)	0.0118 (6.40%)	0.0096 (5.84%)	0.0121 (6.22%)	0.0097 (5.38%)	0.0115 (6.11%)
IOR	0.1758 (100%)	0.1842 (100%)	0.1649 (100%)	0.1947 (100%)	0.1808 (100%)	0.1889 (100%)

Notes: (1) Relative shares are included in parentheses; (2) IOR and all Shapley decompositions are computed using MLD.

presents decomposition results across different redistribution stages, from market income to final disposable income.

The results underscore two key findings. First, parental education accounts for the largest Shapley share of IOR across stages, accounting for roughly half of the explained inequality. Its contribution begins at 9.5% in the market income stage and declines modestly to 7.8% after all redistribution means. This reduction is largely driven by the second distribution, where taxes reduce the contribution by 0.5 percentage points and government transfers by 1.5 percentage points. In contrast, the third redistribution reverses this trend, increasing the contribution by approximately 0.4 percentage points compared to the first distribution. Nonetheless, the consistently high share of IOp associated with parental education affirms its foundational role in relating to economic outcomes—a pattern consistent with findings in other transitional and high-inequality contexts (Bourguignon et al. 2007; Brunori et al. 2013; 2019; Hufe et al. 2022).

Second, and more strikingly, the Shapley share of parental hukou status increases significantly after redistribution. In the pre-tax, pre-transfer stage, it accounts for 5.5% of IOp. This figure rises to 8.3% in the post-redistribution stages. This shift is driven almost entirely by government transfers, which expand hukou’s contribution by 3.2 percentage points. In contrast, taxation slightly reduces hukou’s weight, and inter-household transfers have a negligible effect.

This pattern highlights a critical institutional mechanism: China’s redistributive architecture embeds hukou-based segmentation into the allocation of public benefits. The expansion of hukou’s association power following redistribution suggests that current fiscal instruments—especially contributory pensions—do not neutralize structural disadvantage; they reinforce it. Urban hukou holders are more likely to have formal employment histories and, consequently, greater access to high-benefit schemes such as the Basic Pension Scheme for Urban Employees (BPSUE). These advantages are codified into the benefit formulae of public transfers, effectively translating prior labor-market privilege into post-retirement fiscal reward.

Other circumstance variables—such as gender, ethnicity, parental Communist Party membership, and parental administrative rank—exert relatively minor and stable correlations. Their correlations with IOp remain below 2% in all specifications and show little change across redistribution stages. This consistency suggests that while demographic and political lineage factors may still shape life chances in China, their relative weight is secondary compared to parental educational and hukou-based structural determinants.

Taken together, the Shapley decomposition confirms that IOp in China is structured primarily around two axes: educational inheritance and institutional segmentation via hukou. Redistribution modestly mitigates the former but exacerbates the latter. This duality underscores the importance of evaluating not just whether redistribution occurs, but how it is channeled through systems of eligibility and entitlement that reflect historically embedded inequalities.

These findings add empirical precision to the conceptual claim made earlier in this paper: redistributive policies that rely on institutionally stratified systems—such as contributory pensions—may formally reduce income inequality while substantively increasing the share of inequality that arises from unjust circumstances. In this sense, IOp serves not only as a diagnostic metric but also as a critical lens through which the deeper structure of inequality can be re-evaluated.

Robustness to expanded circumstances is also conducted for the Shapley decomposition. Appendix A.1 Table A3 shows that adding parental occupations (e.g., farmer, employed, self-employed, and chores/other) redistributes Shapley contributions across circumstance dimensions. Because Shapley allocates a fixed pie of explained inequality, introducing new dimensions mechanically reduces the shares previously attributed to existing variables. The qualitative pattern is unchanged: the hukou share remains higher after public transfers than after taxes (baseline: 44.7% vs. 30.3%; expanded: 21.9% vs. 15.4%). At the total stage, the occupation categories together account for roughly 38% of IOR, while the levels for hukou and parental education decline, but their cross-layer ordering persists. Hence, the direction and ordering of channels are robust to a broader definition of inherited circumstances. We therefore emphasize patterns across redistribution layers and rank consistency, not the absolute level of any single share.

In addition, the qualitative ordering of Shapley contributions across income layers is unchanged when the joint parental hukou distribution is balanced, and the parental-hukou share becomes negligible in homogeneous hukou subsamples, as expected. These findings indicate that our conclusions are not driven by sample composition. See Appendix A.1 Table A1.

5. Further Analysis

The previous sections highlighted a central paradox in China’s redistributive structure: while taxation reduces inequality of both outcome and opportunity, government transfers are associated with a rise in inequality of opportunity (IOR increases by 1.89 percentage points after transfers, Table 3). In this section, we isolate the mechanisms within public transfers that account for this pattern and test whether reclassifying pensions as deferred income alters the assessment of redistributive performance.

5.1. Disaggregating Public Transfers: The Role of Pensions

To identify which components of public transfers drive changes in IOp,

Table 5. IOp before and after redistribution within government transfer.

	1st Dis	+Dibao	+Agriculture Subsidies	+Pension	+UI	+Medicare	Public Transfer
IO
MLD	0.5259	0.5186	0.5230	0.5111	0.5249	0.5114	0.5020
Gini	0.5265	0.5242	0.5264	0.5188	0.5260	0.5221	0.5166
IOp
IOA	0.0924 (0.006)	0.0909 (0.005)	0.0909 (0.006)	0.0997 (0.006)	0.0929 (0.005)	0.0914 (0.005)	0.0956 (0.005)
IOR	0.1758	0.1752	0.1738	0.1950	0.1769	0.1786	0.1947
Gender	0.0066	0.0066	0.0070	0.0050	0.0067	0.0064	0.0055
PE	0.0946	0.0937	0.0947	0.0823	0.0942	0.0917	0.0793
PH	0.0549	0.0555	0.0528	0.0848	0.0564	0.0600	0.0871

Note: (1) Bootstrap standard errors are in parentheses; (2) Gini is reported for comparability; IOp metrics (IOA and IOR) and all Shapley decompositions are computed using MLD; (3) PE stands for Parental Education, PH stands for Parental Hukou. Data Source: CHFS 2013 and authors’ calculation.

reports inequality measures after separately adding each major transfer program—minimum living standard assistance (Dibao), agricultural subsidies, pensions, unemployment insurance, and medical reimbursements—to market income. Among these, pensions exhibit the largest and most regressive effect on IOp, raising the IOR from 0.1758 (market income) to 0.1950—an increase of 1.92 percentage points. This is more than the total effect of all public transfers combined. This outcome likely stems from the household-based allocation of pensions, which makes them strongly correlated with household economic status (Cai and Yue 2020). This outcome is explained by both a rise in IOA and a decline in IO, indicating that pension benefits disproportionately accrue to those with favorable circumstantial backgrounds. The rise in IOR can be attributed to an increase in IOA and a decrease in IO, reflecting the dual nature of the contributory pension system. Although retirees (classified as non-labor force) receive compensation to narrow the income gap with the active workforce, institutional barriers within the pension system, particularly those arising from hukou-based segregation, further exacerbate IOp.

A further decomposition of circumstantial factors reveals that the difference in parental hukou status is the primary contributor to the exacerbating effect of pensions on IOp. In the Shapley accounting, while pension reduces the contribution of parental education to IOp by approximately 1 percentage point, it increases the contribution from parental hukou by approximately 3 percentage points. This may be attributed to the distinct pension schemes for urban and rural residents in China, which will be further discussed in the following section.

In contrast, other transfers display relatively modest effects on IOp. Agricultural subsidies and Dibao each reduce IOR slightly (by 0.1–0.2 percentage points), while unemployment and health insurance reimbursements raise IOR marginally (by 0.1–0.3 points). The minimal influence of these programs reflects their narrow coverage and relatively low benefit levels.

Taken together, these findings suggest that the regressive effect of public transfers on IOp is driven almost entirely by pensions, while other transfer components are either neutral or mildly progressive. This observation elevates pensions from one among many redistributive instruments to the central institutional lever shaping IOp through the second distribution.

5.2. A Classification Problem: Are Pensions Transfers or Deferred Income?

To understand the mechanism through which pensions increase IOp while reducing overall income inequality (IO) in China, it is necessary to briefly examine the structure of the pension system and its function as a redistributive instrument.

Pension systems are broadly classified into two conceptual models: Pension as Deferred Income (PDI) and Pension as a Government Transfer (PGT). The PDI model views pensions as earned entitlements, in which retirement benefits represent deferred wages accrued through an individual’s previous labor market participation and financial contributions (Lustig 2018). Under this model, pension benefits are closely linked to earnings history and contribution levels, reinforcing a direct relationship between past employment and post-retirement income. In contrast, the PGT model treats pensions as state-financed transfers, providing benefits independently of an individual’s contribution history and functioning as a social safety net to ensure a minimum standard of living for retirees. This conceptual distinction offers valuable insight into the ongoing debate over whether pensions should be regarded as deferred market income (Breceda et al. 2009; Lustig 2011) or as government transfer payments (Lindert et al. 2006; Goñi et al. 2011).

China’s pension system predominantly follows the PDI model, where retirement benefits are largely determined by individual and employer contributions, rather than by universally funded government transfers. The system comprises two primary schemes: the Basic Pension Scheme for Urban Employees (BPSUE) and the Basic Pension Scheme for Rural and Non-Working Urban Residents (BPSRNUR). Although these schemes differ in terms of eligibility, contribution requirements, and benefit structures, both are fundamentally contributory. This reinforces the view that pensions in China are primarily earned through financial participation rather than granted as a universal entitlement (CICC Research 2024).

The Basic Pension Scheme for Urban Employees (BPSUE)20 serves as the primary pension system for individuals employed in state-owned enterprises (SOEs), private enterprises, and self-employed sectors21. This scheme operates on a pay-as-you-go (PAYG) basis, in which current workers contribute to a pension pool that finances retirees’ benefits while simultaneously accruing entitlements for their own future pensions.

BPSUE is financed through payroll contributions, with employers contributing 20% of an employee’s wages and employees contributing 8% to their individual pension accounts. Pension benefits are directly linked to prior earnings and the duration of contributions, with an estimated replacement rate of 50–60% (Fang and Feng 2020). Because benefits under this scheme are entirely derived from prior financial contributions, BPSUE aligns fully with the PDI framework, framing pensions as deferred wages rather than redistributive transfers.

The Basic Pension Scheme for Rural and Non-Working Urban Residents (BPSRNUR) was established to extend pension coverage to individuals outside formal urban employment, including rural laborers and non-salaried urban residents. Unlike BPSUE, participation in BPSRNUR is voluntary. Participants may choose to contribute between 100 and 2000 RMB annually, which is deposited directly into individual accounts—functioning similarly to a personal savings account (Liu and Sun 2016). The scheme’s voluntary participation and flexible contribution levels suggest that pension benefits are partially determined by individual saving behavior, rather than being provided as a universal entitlement. To some extent, this can be interpreted as a form of deferred income or a quasi-mandatory savings mechanism. It is also important to note that BPSRNUR incorporates a government transfer component, as the pooling account is subsidized by the state22 rather than relying solely on participant contributions (CICC Research 2024). As government subsidies supplement individual savings, some view BPSRNUR functions as a hybrid between the PDI and PGT models—where part of the benefits derives from personal contributions (deferred income), and the remainder is provided through state-financed transfers.

However, we argue that BPSRNUR aligns more with the PDI than the PGT model due to its contributory nature, individual accounts, and hybrid funding model. Unlike purely redistributive transfers, BPSRNUR requires contributions that accumulate over time, linking benefits directly to prior earnings rather than need-based allocation (CICC Research 2024). Individual accounts further incentivize long-term participation, reinforcing the pension’s role as a form of delayed compensation rather than an outright government subsidy (Chang et al. 2024). Additionally, while government support supplements the scheme, funding is also derived from contributions, distinguishing it from tax-financed welfare programs. These features position BPSRNUR as a system that defers income for retirement security rather than merely redistributing wealth.

Therefore, the assumption that pension is viewed as PDI and should be categorized in the first distribution may be a plausible explanation of the mechanism that raises IOp while reducing IO.

Table 6. IOp before and after the redistribution when reclassified pension.

	1st Dis	1st Dis (w/Pension)	2nd Dis (w/o Pension)	Total (w/o Pension)	Total (w/Pension)
IO
MLD	0.5259	0.5111	0.4875	0.4797	0.4706
Gini	0.5265	0.5188	0.5110	0.5091	0.5264
IOp
IOA	0.0924 (0.006)	0.0997 (0.006)	0.0816 (0.005)	0.0826 (0.005)	0.0889 (0.005)
IOR	0.1758	0.1950	0.1673	0.1722	0.1889
Gender	0.0066	0.0050	0.0075	0.0068	0.0053
PE	0.0946	0.0823	0.0856	0.0899	0.0782
PH	0.0549	0.0848	0.0551	0.0565	0.0833

Note: (1) Bootstrap standard errors are in parentheses; (2) Gini is reported for comparability; IOp metrics (IOA and IOR) and all Shapley decompositions are computed using MLD. Data Source: CHFS 2013 and authors’ calculation.

further explores this assumption by reclassifying pension from a component of government transfers (second distribution) to deferred market income (first distribution). Different from the baseline results, the second distribution reduces both IO and IOp when the pension is reclassified as the first distribution. Specifically, the IOR decreases by nearly 3 percentage points. Aligning with previous analysis, the reduction in IOR results from the reduction in both IO and IOA, where the reduction in the latter is smaller. The reduction in IOR is mainly contributed to by parental hukou status, as the institutional barriers in the pension system (i.e., hukou segregation) are among the key contributors to IOp in China (Hung 2022).

Given the complex nature of pensions, an alternative approach is to assess the redistributive effect of the second distribution while excluding pensions from the analysis. Under this alternative classification, the second distribution also reduces the IOR by nearly 1 percentage point. This finding highlights the effectiveness of second distribution mechanisms in reducing income inequality in China, both from outcome and opportunity perspectives, when pensions are excluded.

6. Regional Heterogeneity

We assess whether the main patterns differ across eastern, central, and western China. The headline result holds: taxation is modestly equalizing everywhere, whereas the second distribution appears regressive for IOp only when pensions are included, with the strongest effect in the east and center and muted in the west. In the east and center, pensions alone account for nearly the entire rise in IOR after transfers, consistent with higher formal-employment coverage and benefit generosity tied to urban hukou histories. In the West, where formalization and urban-scheme coverage are lower, transfers are closer to neutral, and the net effect is small. The third distribution shows weak, slightly regressive associations across regions, consistent with unequal private support capacities. Shapley attribution indicates that parental education remains the largest contributor to IOp in all regions, while parental hukou’s share increases after transfers, again tracing the pension channel. Full regional estimates, standard errors, and variable contributions are reported in Appendix A.2,

Table A4. Regional Heterogeneity in IOps before and after Redistribution.

	Eastern			Middle			Western
	IOR	PE	PH	IOR	PE	PH	IOR	PE	PH
1st	0.1865 (0.009)	0.1034 (55.43)	0.0525 (28.15)	0.1418 (0.006)	0.0590 (41.63)	0.0543 (38.27)	0.2267 (0.010)	0.120 (52.78)	0.0683 (30.12)
2nd	0.1980 (0.009)	0.0825 (41.66)	0.0840 (42.41)	0.1547 (0.007)	0.0405 (26.20)	0.0838 (54.19)	0.2222 (0.010)	0.0957 (43.09)	0.0858 (38.60)
#Tax	0.1790 (0.008)	0.0985 (55.07)	0.0500 (27.95)	0.1251 (0.006)	0.0532 (42.50)	0.0451 (36.03)	0.2145 (0.010)	0.1131 (52.72)	0.0628 (29.26)
#PT	0.2060 (0.010)	0.0878 (42.65)	0.0861 (41.81)	0.1711 (0.006)	0.0457 (26.73)	0.0933 (54.56)	0.2342 (0.010)	0.1018 (43.49)	0.0918 (39.23)
#Pension	0.2052 (0.009)	0.0881 (42.93)	0.0852 (41.53)	0.1694 (0.007)	0.0488 (28.78)	0.0897 (52.96)	0.2375 (0.009)	0.1097 (46.19)	0.0868 (36.56)
3rd	0.1941 (0.007)	0.1093 (56.32)	0.0548 (28.24)	0.1440 (0.005)	0.0614 (42.64)	0.0552 (38.31)	0.2278 (0.011)	0.1220 (53.58)	0.0663 (29.12)
Total	0.2055 (0.007)	0.0880 (42.84)	0.0868 (42.24)	0.1555 (0.006)	0.0434 (27.93)	0.0827 (53.20)	0.2240 (0.011)	0.0991 (44.23)	0.0833 (37.18)

Notes: (1) Bootstrap standard errors are in parentheses under IORs; (2) relative contributions to IORs are included in parentheses under PE and PH.

(detailed discussion and interpretation are also included in Appendix A.2). These results confirm that our baseline mechanism—pension-linked institutional segmentation—is not a single-region artifact but varies in intensity with local labor-market formalization and scheme coverage.

7. Conclusions

This study investigated the redistributive structure of income inequality in China through the lens of inequality of opportunity (IOp). While prior research has examined how taxes and transfers affect overall income distribution, relatively little is known about whether and how these mechanisms reduce unfair disparities rooted in structural circumstances, such as family background, hukou status, and educational inheritance. Motivated by this gap, and building on normative frameworks distinguishing outcomes from opportunities, this paper asked: Does redistribution in China equalize opportunity, or does it reinforce existing structural advantages?

Using nationally representative data from the 2013 China Household Finance Survey and an ex-ante parametric estimation of IOp, we presented four main findings.

First, while China’s redistributive system reduces total income inequality, it increases the relative contribution of circumstances to inequality, thereby intensifying inequality of opportunity by more than one percentage point after all redistribution means. That is, redistribution compresses income gaps but leaves the structural roots of inequality unaddressed—and, in some cases, further entrenched. This finding aligns with existing literature, which emphasizes that redistributive mechanisms often prioritize reducing income inequality of outcome (IO) rather than directly addressing disparities stemming from exogenous circumstances (Roemer et al. 2003). This suggests that the current redistribution mechanism may be insufficient in mitigating disparities in opportunity, especially if policies fail to account for structurally disadvantaged groups.

Second, this regressive shift in IOp is not driven by taxation, which reduces both IO and IOp. Rather, it is government transfers—particularly pensions—that disproportionately benefit individuals with advantageous backgrounds. These transfers amplify the contribution of parental hukou status to IOp, a finding consistent with institutional accounts of China’s welfare segmentation (Wu and Zheng 2018; Zhou and Song 2016). Although hukou conversion is formally possible, initial hukou classification is assigned at birth and shaped by parental status—placing it squarely within the realm of circumstance (Zhang and Treiman 2013). Government transfers thus reinforce, rather than repair, inherited inequalities.

This result is particularly pronounced in the case of pensions, which significantly increases IOp despite reducing overall income inequality. While pension schemes provide critical income support for retirees, their design is structurally biased: access and benefit levels are tightly linked to formal employment histories, which themselves depend on education, urban registration, and elite employment sectors. As a result, pensions reproduce labor market inequalities and institutional stratification, a pattern well-documented in prior research (Zhu and Walker 2018). This study confirms empirically that pensions in China function less as redistributive transfers and more as institutionally gated deferred income.

Third, inter-household transfers—comprising the third distribution—also exacerbate IOp, albeit to a lesser degree. These transfers raise IOR by approximately 0.5 percentage points, revealing that familial and kinship networks play a non-trivial role in reproducing advantage (Corak 2013; Swartz 2009). While such transfers are often invisible in fiscal accounts, they reflect private capacities to mobilize resources that vary widely by household background. In practice, wealthier families are better equipped to provide educational, housing, and emergency support, further widening opportunity gaps.

Fourth, we show that the redistributive effect of pensions on IOp is highly sensitive to classification. When pensions are treated as public transfers, redistribution appears regressive from an IOp perspective. However, when reclassified as deferred income, consistent with their contributory logic, the increase in IOp is largely neutralized. This reclassification substantially reduces the apparent regressivity of redistribution, especially the contribution of parental hukou to explained inequality. The finding highlights the critical importance of institutional assumptions in evaluating equity.

The study offers several policy-relevant insights: First, pension integration and reform should be a priority. Equalizing access to public pensions across formal and informal sectors—particularly through the harmonization of urban and rural schemes—would reduce the opportunity-reinforcing effects currently embedded in benefit eligibility rules. Second, rebalancing redistribution toward universal or means-tested programs can enhance equity. While contributory pensions track labor market history, non-contributory transfers can be designed to explicitly target opportunity-deprived groups, such as rural, migrant, or low-education households. Third, reducing the redistributive burden on private transfers is essential. Informal support networks are unevenly distributed and often amplify socioeconomic gaps. Expanding formal support for housing, education, and caregiving would reduce reliance on family-based redistribution and help neutralize intergenerational privilege. Fourth, IOp metrics should be embedded in policy evaluation frameworks. Traditional income-based assessments may overlook whether redistribution genuinely closes structural gaps. Institutionalizing IOp analysis would offer a more ethically grounded and analytically precise benchmark for evaluating redistributive performance.

The study has three limitations. First, as stated and justified before, the use of 2013 data is indeed outdated for the current redistributive effectiveness assessment. Second, it is based on cross-sectional data and does not capture intertemporal dynamics of mobility or IOp. Future work using panel data could explore how opportunity (dis)advantage evolves across life stages. Third, while the study infers effort through residual income variance, future research could incorporate richer proxies of effort or ability to better isolate structural disadvantages.

Nevertheless, the central conclusion remains robust: Redistribution in China reduces inequality of outcome but not necessarily inequality of opportunity. Unless institutional segmentation is addressed and redistribution mechanisms are restructured, fiscal policy will remain a limited tool for promoting intergenerational fairness. This study highlights the need to move beyond equalizing outcomes toward building redistributive systems that dismantle the structural origins of inequality itself.