pith. sign in

arxiv: 2605.17768 · v1 · pith:P5CM7W3Xnew · submitted 2026-05-18 · 💱 q-fin.RM

Mortality Heterogeneity and Actuarial Fairness in China's Notional Defined Contribution Pension System

Xiaoyu Dong , Hong Li , Kenneth Q. Zhou , Xiaobai Zhu This is my paper

Pith reviewed 2026-05-20 00:57 UTC · model grok-4.3

classification 💱 q-fin.RM
keywords actuarial fairnessNDC pension systemmortality heterogeneityincome groupsannuity divisorChina pensionLee-Carter modelreverse transfer
0
0 comments X

The pith

China's age-only pension divisor subsidizes higher-income retirees due to mortality differences across groups.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper examines actuarial fairness in China's notional defined contribution pension system when life expectancies vary by income. Current rules convert account balances to benefits using a divisor that depends only on retirement age, without adjusting for income-linked mortality patterns. The authors build a model that combines national mortality records with survey subgroup data and uses Hermite splines to estimate separate baseline death rates for each income group while sharing a common time trend. Applying this setup shows that higher earners receive benefits whose present value exceeds their contributions by a larger margin than lower earners do, producing both an overall shortfall and a transfer of value from poorer to richer retirees. The study also tests four practical income-adjusted divisor designs and finds each one cuts the size of that reverse transfer.

Core claim

When mortality schedules are allowed to differ by income in a Lee-Carter framework that keeps a shared period effect but lets baseline schedules vary, the official age-only annuity divisor used in China's NDC system generates substantial actuarial unfairness. The subsidy to each retiree rises steadily with income, which implies both an aggregate shortfall for the system and a net transfer of resources from lower-income to higher-income retirees. Four implementable alternatives that make the divisor depend on income all reduce the magnitude of this reverse transfer.

What carries the argument

Mortality-differentiated Lee-Carter model with group-specific baseline mortality schedules parameterized by Hermite splines and a common period effect.

Load-bearing premise

The Hermite spline parameterization of group-specific baseline mortality schedules estimated from limited survey subgroup data combined with national aggregates accurately captures true income-related mortality differences without material bias or extrapolation error.

What would settle it

Recompute the subsidies using a new, larger dataset of actual deaths and incomes by group over the same or later years; if the monotonic rise in subsidy with income disappears or reverses, the central claim is falsified.

Figures

Figures reproduced from arXiv: 2605.17768 by Hong Li, Kenneth Q. Zhou, Xiaobai Zhu, Xiaoyu Dong.

Figure 1
Figure 1. Figure 1: Basis functions of Hermite splines over standardized age ˜x [PITH_FULL_IMAGE:figures/full_fig_p008_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Log mortality curves of the fitted Gompertz models by income quintile. [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Summary of the CHARLS dataset used for estimation. [PITH_FULL_IMAGE:figures/full_fig_p014_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Lee–Carter parameters fitted to China’s national mortality data. [PITH_FULL_IMAGE:figures/full_fig_p015_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Log mortality curves of the fitted HSM-III baseline model by income quintile. [PITH_FULL_IMAGE:figures/full_fig_p018_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Actuarially fair counting month Mfair x,j,t (top row) and subsidy rate Λx,j,t (bottom row) at years t = 2020 − 2040 by income quintile. 20 [PITH_FULL_IMAGE:figures/full_fig_p020_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Comparison of the four pension conversion methods across career-average income [PITH_FULL_IMAGE:figures/full_fig_p034_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Residual subsidy rate after applying each of the four methods. [PITH_FULL_IMAGE:figures/full_fig_p035_8.png] view at source ↗
read the original abstract

We study actuarial fairness in China's notional defined contribution (NDC) pension system when mortality differs across income groups. Under current rules, individual account balances are converted into monthly benefits using an official annuity divisor that depends only on retirement age. We develop a mortality-differentiated Lee-Carter framework with group-specific baseline mortality schedules and a common period effect, estimated by combining national mortality data for 1994-2020 with CHARLS subgroup data for 2011-2020. To model cross-group mortality parsimoniously under limited data, we parameterize the baseline schedules using Hermite splines. Applying the model to China's NDC system, we find substantial actuarial unfairness in the current age-only divisor. The subsidy rises monotonically with income, implying both an aggregate actuarial shortfall and a reverse transfer from poorer to richer retirees. We then compare four implementable income-dependent annuitization rules, ranging from a simple bracket design to marginal-rule alternatives, and show that all substantially reduce the reverse transfer.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper develops a Lee-Carter mortality model with income-group-specific baseline schedules parameterized by Hermite splines, estimated by combining national Chinese mortality aggregates (1994-2020) with CHARLS income-stratified observations (2011-2020). It applies the fitted model to China's NDC pension rules, showing that the current age-only annuity divisor produces substantial actuarial unfairness: subsidies increase monotonically with income, generating both an aggregate shortfall and a reverse transfer from poorer to richer retirees. The authors then compare four implementable income-dependent annuitization rules and report that all reduce the reverse transfer.

Significance. If the estimated income-mortality gradients hold, the results quantify a concrete equity problem in an important pension system and supply practical, bracket-based or marginal-rule fixes that could be implemented with existing administrative data. The work contributes to the actuarial literature on heterogeneity-adjusted annuities and supplies policy-relevant evidence for NDC reforms in aging economies.

major comments (2)
  1. [§4.2] §4.2 (Estimation): The Hermite spline parameterization of group-specific baselines, fitted to limited CHARLS subgroup cells, is load-bearing for the monotonic subsidy result; small effective sample sizes per income stratum and possible selection into the survey can bias the older-age mortality schedules that determine the annuity divisors, directly scaling the reported subsidies in §5.2.
  2. [§5.1] §5.1, Table 5: The claim of monotonic subsidy increase with income and the resulting reverse transfer rest on the projected group-specific mortality schedules; without reported sensitivity checks to alternative knot placements, extrapolation methods, or period-effect specifications, the central fairness conclusion remains sensitive to the weakest modeling assumption.
minor comments (2)
  1. [§3.1] §3.1: The notation for the common period effect κ_t versus the group-specific baselines μ_x^{(g)} could be introduced earlier to improve readability of the model equations.
  2. [Figure 4] Figure 4: Axis labels on the projected mortality schedules should explicitly note the income-group ordering to match the text description.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments on our manuscript. We address each major comment below and indicate the revisions we will undertake to strengthen the analysis.

read point-by-point responses

  1. Referee: [§4.2] §4.2 (Estimation): The Hermite spline parameterization of group-specific baselines, fitted to limited CHARLS subgroup cells, is load-bearing for the monotonic subsidy result; small effective sample sizes per income stratum and possible selection into the survey can bias the older-age mortality schedules that determine the annuity divisors, directly scaling the reported subsidies in §5.2.

    Authors: We acknowledge that the CHARLS sample sizes decline sharply at older ages within each income stratum and that survey participation may introduce selection effects. The Hermite spline specification was adopted precisely to impose smoothness and limit the number of free parameters given these data constraints, while the common period effect is identified from the much larger national mortality aggregates (1994–2020). In the revision we will add a table reporting effective cell sizes by age and income group, together with a brief discussion of how the spline penalty and national data anchor the older-age schedules. We will also include a short robustness section comparing results under alternative spline knot placements. revision: yes

  2. Referee: [§5.1] §5.1, Table 5: The claim of monotonic subsidy increase with income and the resulting reverse transfer rest on the projected group-specific mortality schedules; without reported sensitivity checks to alternative knot placements, extrapolation methods, or period-effect specifications, the central fairness conclusion remains sensitive to the weakest modeling assumption.

    Authors: We agree that the manuscript would benefit from explicit sensitivity checks. Although the monotonic subsidy pattern is stable across the specifications we examined internally, we did not report these checks. In the revised manuscript we will add an appendix that varies (i) the number and location of Hermite spline knots, (ii) the extrapolation rule for ages beyond the observed CHARLS range, and (iii) the functional form of the period effect. We will show that the qualitative conclusions—substantial reverse transfers under the current age-only divisor and their reduction under income-dependent rules—remain intact under these perturbations. revision: yes

Circularity Check

0 steps flagged

No circularity: unfairness finding is direct application of externally estimated mortality schedules to fixed external rules

full rationale

The paper estimates group-specific baseline mortality via a Lee-Carter structure with Hermite-spline parameterization, using national 1994-2020 aggregates plus CHARLS 2011-2020 income-stratified observations. These fitted schedules are then inserted into the official age-only annuity divisor formulas (which are policy parameters external to the model) to compute actuarial subsidies and transfers. The monotonic subsidy-with-income result is therefore an output of the estimated mortality differentials applied to independent rules, not a quantity defined by or forced to equal the fitted parameters themselves. No self-citation chain, uniqueness theorem, or ansatz smuggling is invoked to justify the central claim, and the derivation remains falsifiable against the underlying mortality data.

Axiom & Free-Parameter Ledger

2 free parameters · 1 axioms · 0 invented entities

The central claim rests on fitted spline and period-effect parameters estimated from external mortality data plus the standard Lee-Carter structural assumption; no new entities are introduced.

free parameters (2)
  • Hermite spline coefficients for group-specific baselines
    Chosen to fit limited subgroup data parsimoniously.
  • Lee-Carter period effect parameters
    Estimated jointly from national and CHARLS data.
axioms (1)
  • domain assumption Mortality follows Lee-Carter form with group-specific age schedules and common period effect.
    Invoked to enable parsimonious modeling of heterogeneity given data constraints.

pith-pipeline@v0.9.0 · 5712 in / 1384 out tokens · 82401 ms · 2026-05-20T00:57:01.065473+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

37 extracted references · 37 canonical work pages

  1. [1]

    Bai, R., Liu, Y., Zhang, L., Dong, W., Bai, Z., and Zhou, M. (2023). Projections of future life expectancy in China up to 2035: a modelling study.The Lancet Public Health, 8(12):e915–e922

    work page 2023

  2. [2]

    Brown, J. R. (2001). Private pensions, mortality risk, and the decision to annuitize.Journal of public Economics, 82(1):29–62

    work page 2001

  3. [3]

    Brown, J. R. (2003). Redistribution and insurance: Mandatory annuitization with mortality het- erogeneity.Journal of Risk and insurance, 70(1):17–41. 38

    work page 2003

  4. [4]

    J., Blake, D., and Dowd, K

    Cairns, A. J., Blake, D., and Dowd, K. (2006). A two-factor model for stochastic mortality with parameter uncertainty: theory and calibration.Journal of Risk and Insurance, 73(4):687–718

    work page 2006

  5. [5]

    J., Blake, D., Dowd, K., Coughlan, G

    Cairns, A. J., Blake, D., Dowd, K., Coughlan, G. D., Epstein, D., Ong, A., and Balevich, I. (2009). A quantitative comparison of stochastic mortality models using data from England and Wales and the United States.North American Actuarial Journal, 13(1):1–35

    work page 2009

  6. [6]

    J., Kallestrup-Lamb, M., Rosenskjold, C

    Cairns, A. J., Kallestrup-Lamb, M., Rosenskjold, C. P. T., Blake, D., and Dowd, K. (2016). Mod- elling socio-economic differences in the mortality of Danish males using a new affluence index

    work page 2016

  7. [7]

    R., and Diamond, P

    Davidoff, T., Brown, J. R., and Diamond, P. A. (2005). Annuities and individual welfare.American Economic Review, 95(5):1573–1590. del Carmen Boado-Penas, M., Haberman, S., and Naka, P. (2022). Fairness and annuity divisors for notional defined contribution pension schemes.Journal of Pension Economics & Finance, 21(2):143–167

    work page 2005

  8. [8]

    and Blake, D

    Dowd, K. and Blake, D. (2022). Projecting mortality rates to extreme old age with the CBDX model.Forecasting, 4(1):208–218

    work page 2022

  9. [9]

    and Feng, J

    Fang, H. and Feng, J. (2018). The chinese pension system. Technical report, National Bureau of Economic Research

    work page 2018

  10. [10]

    H., Mitchell, O

    Fong, J. H., Mitchell, O. S., and Koh, B. S. (2011). Longevity risk management in singapore’s national pension system.Journal of Risk and Insurance, 78(4):961–982

    work page 2011

  11. [11]

    Fritsch, F. N. and Carlson, R. E. (1980). Monotone piecewise cubic interpolation.SIAM Journal on Numerical Analysis, 17(2):238–246

    work page 1980

  12. [12]

    Gavrilov, L. A. and Gavrilova, N. S. (2005). Reliability theory of aging and longevity.Handbook of the Biology of Aging, pages 3–42

    work page 2005

  13. [13]

    and Webb, A

    Gong, G. and Webb, A. (2008). Mortality heterogeneity and the distributional consequences of mandatory annuitization.Journal of Risk and Insurance, 75(4):1055–1079. 39

    work page 2008

  14. [14]

    Horneff, V., Maurer, R., and Mitchell, O. S. (2023). Fixed and variable longevity income annuities in defined contribution plans: Optimal retirement portfolios taking social security into account. Journal of Risk and Insurance, 90(4):831–860

    work page 2023

  15. [15]

    Huang, F., Hui, F., and Villegas, A. (2025). Towards fairer retirement outcomes: Socio-economic mortality differentials in australia.Available at SSRN 5253598

    work page 2025

  16. [16]

    and Blake, D

    Hunt, A. and Blake, D. (2021). On the structure and classification of mortality models.North American Actuarial Journal, 25(sup1):S215–S234

    work page 2021

  17. [17]

    Hyman, J. M. (1983). Accurate monotonicity preserving cubic interpolation.SIAM Journal on Scientific and Statistical Computing, 4(4):645–654

    work page 1983

  18. [18]

    Jijiie, A., Alonso-Garc´ ıa, J., and Arnold, S. (2022). Mortality by socio-economic class and its impact on the retirement schemes: How to render the systems fairer?European Actuarial Journal, 12(2):701–743

    work page 2022

  19. [19]

    Kochanek, D. H. and Bartels, R. H. (1984). Interpolating splines with local tension, continuity, and bias control. InProceedings of the 11th annual conference on Computer graphics and interactive techniques, pages 33–41

    work page 1984

  20. [20]

    Lee, R. D. and Carter, L. R. (1992). Modeling and forecasting US mortality.Journal of the American statistical association, 87(419):659–671

    work page 1992

  21. [21]

    and Niu, J

    Li, D. and Niu, J. (2024). The gender gap in pension income for urban employees in China.Sage Open, 14(3):21582440241266996

    work page 2024

  22. [22]

    and Yi, F

    Li, D. and Yi, F. (2025). The gender differences in pension wealth of individual retirement accounts in China.Journal of Economic Statistics, 3(1):17–31

    work page 2025

  23. [23]

    S.-H., Zhou, K

    Li, J. S.-H., Zhou, K. Q., Zhu, X., and Chan, W.-S. (2026). A national-provincial stochastic mortality model for china: Methods, innovations, and policy applications.Available at SSRN 6212720

    work page 2026

  24. [24]

    S.-H., Zhou, K

    Li, J. S.-H., Zhou, K. Q., Zhu, X., Chan, W.-S., and Chan, F. W.-H. (2019). A bayesian approach 40 to developing a stochastic mortality model for china.Journal of the Royal Statistical Society Series A: Statistics in Society, 182(4):1523–1560

    work page 2019

  25. [25]

    and Lee, R

    Li, N. and Lee, R. D. (2005). Coherent mortality forecasts for a group of populations: An extension of the Lee-Carter method.Demography, 42(3):575–594

    work page 2005

  26. [26]

    and Dandapani, K

    Lu, X. and Dandapani, K. (2023). Design of employee pension benefits model and China’s pension gender gap.Global Finance Journal, 56:100828

    work page 2023

  27. [27]

    Moore, J. C. and Welniak, E. J. (2000). Income measurement error in surveys: A review.Journal of official statistics, 16(4):331

    work page 2000

  28. [28]

    and de Gosson de Varennes, Y

    Palmer, E. and de Gosson de Varennes, Y. Z. (2019).Annuities in (N) DC Pension Schemes: Design, heterogeneity, and estimation issues. World Bank Washington, DC

    work page 2019

  29. [29]

    Peng, W., Lin, S., Chen, B., Bai, X., Wu, C., Zhang, X., Yang, Y., Cui, J., Xu, W., Song, L., et al. (2024). Area-level socioeconomic inequalities in mortality in China: a nationwide cohort study based on the chinaheart project.The Lancet Public Health, 9(12):e1014–e1024

    work page 2024

  30. [30]

    Richards, S. J. (2020). A Hermite-spline model of post-retirement mortality.Scandinavian Actuarial Journal, 2020(2):110–127

    work page 2020

  31. [31]

    Shen, K., Wang, F., and Cai, Y. (2016). Patterns of inequalities in public transfers by gender in China.The Journal of the Economics of Ageing, 8:76–84

    work page 2016

  32. [32]

    Strehler, B. L. and Mildvan, A. S. (1960). General theory of mortality and aging: A stochas- tic model relates observations on aging, physiologic decline, mortality, and radiation.Science, 132(3418):14–21

    work page 1960

  33. [33]

    Tang, S., Li, J., and Tickle, L. (2023). A hermite spline approach for modelling population mortality. Annals of Actuarial Science, 17(2):243–284

    work page 2023

  34. [34]

    Weitz, J. S. and Fraser, H. B. (2001). Explaining mortality rate plateaus.Proceedings of the national academy of sciences, 98(26):15383–15386

    work page 2001

  35. [35]

    Yaari, M. E. (1965). Uncertain lifetime, life insurance, and the theory of the consumer.The Review of Economic Studies, 32(2):137–150. 41

    work page 1965

  36. [36]

    Zhang, J., Zhu, X., and Wei, W. (2025). Welfare-enhancing annuity divisor for notional defined contribution design.Insurance: Mathematics and Economics, page 103191

    work page 2025

  37. [37]

    A., Zhao, Y., Liu, Z., Wang, C., Xiang, Q., Rangarajan, S., Li, S., et al

    Zhu, Y., Wang, Y., Shrikant, B., Tse, L. A., Zhao, Y., Liu, Z., Wang, C., Xiang, Q., Rangarajan, S., Li, S., et al. (2023). Socioeconomic disparity in mortality and the burden of cardiovascular disease: analysis of the Prospective Urban Rural Epidemiology (PURE)-China cohort study.The Lancet Public Health, 8(12):e968–e977. A Official Counting Months Age (...

    work page 2023