pith. sign in

arxiv: 2606.27924 · v1 · pith:LC4N5FP4new · submitted 2026-06-26 · 💰 econ.GN · q-fin.EC

Heterogeneous Diffusion of Electric Vehicles in China: Demand, Learning, Product Entry, and the Incidence of Industrial Policy

Yu (Jasmine) Hao , Jinge Li This is my paper

Pith reviewed 2026-06-29 02:16 UTC · model grok-4.3

classification 💰 econ.GN q-fin.EC
keywords electric vehiclesChina auto marketindustrial policysubsidy welfareWright's law learningShapley decompositiontechnology adoptiondemand estimation
0
0 comments X

The pith

Every yuan of 2024 EV subsidy in China delivered 3.38 yuan of private surplus, with learning effects accounting for half the welfare impact of removal.

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

The paper uses an equilibrium differentiated-products model of the Chinese auto market to attribute the 2015-2024 rise in EV sales share from 1% to 45%. It decomposes that rise into six channels and measures the welfare consequences of subsidies, including their indirect effects on costs via learning. The central finding is that product-quality gains, choice-set expansion, and battery-cost declines explain most of the adoption increase, while each subsidy yuan returns more than three yuan in private surplus but with strong regional and firm-type heterogeneity. About half the aggregate welfare loss from removing the 2024 subsidy operates through reduced future learning rather than the direct transfer. Readers care because the results separate direct policy effects from dynamic supply-side responses in a large-scale technology transition.

Core claim

An equilibrium differentiated-products model of the Chinese auto market shows that the 2015-2024 EV share rise is driven primarily by product-quality gains (+45.49%), choice-set expansion (+14.81%), and battery-cost decline (+8.20%), with the subsidy channel contributing negatively (-13.63%) because of phase-down. Every yuan of 2024 EV subsidy generated 3.38 yuan of private surplus, half of the welfare loss from subsidy removal occurs through indirect Wright's-law learning, per-capita consumer-surplus loss is five times larger in Tier-1 cities than in the Rest tier, and EV-native firms retain 16-27% of 2024 EV business under removal while traditional state-owned firms retain only 11%. A sepa

What carries the argument

Equilibrium differentiated-products demand model with Shapley-value decomposition across Quality, Variety, Battery, Subsidy, Residual, and Market channels, combined with a Wright's-law cost function.

If this is right

  • Subsidy removal reduces 2024 EV share by 23-33%.
  • Consumer-surplus loss from removal is five times larger per capita in Tier-1 cities than in lower tiers.
  • EV-native firms retain 16-27% of their EV business after removal while state-owned manufacturers retain 11%.
  • Roughly half the aggregate welfare loss from subsidy removal operates through the indirect learning channel.
  • The subsidy block contributes negatively to historical share growth because direct purchase subsidies were phased down.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Policies that directly expand model variety or accelerate quality improvements could raise EV adoption more efficiently than equivalent spending on purchase subsidies.
  • The large regional asymmetry implies that uniform national subsidies may be less cost-effective than tier-targeted designs.
  • The same learning-channel logic could be tested in other countries or technologies where cost declines depend on cumulative production.
  • Post-2024 data on EV sales and battery prices after subsidy changes would provide an out-of-sample check on the model's learning parameters.

Load-bearing premise

The demand system, supply-side pricing, and Wright's-law cost function are correctly specified and identified so that the Shapley decomposition and welfare numbers recover true marginal contributions.

What would settle it

Whether removing the 2024 subsidy produces the predicted 23-33% drop in EV share or whether post-2024 battery cost changes match the estimated learning rate from the Wright's-law function.

Figures

Figures reproduced from arXiv: 2606.27924 by Jinge Li, Yu (Jasmine) Hao.

Figure 1
Figure 1. Figure 1: Rebuilt national NEV subsidy path. Notes: Cash subsidies decline in steps and end after 2022; the 2023–2024 direct subsidy is mainly the purchase-tax exemption. through lottery or auction. In Shanghai, an auctioned ICE plate trades for roughly 90,000 yuan, and EVs are exempt from the auction. The implicit value of a free EV plate is comparable to the early-period cash subsidy in those cities. We do not inc… view at source ↗
Figure 2
Figure 2. Figure 2: EV and GV attribute progress, 2015–2024. [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Recovered BLP mean utility 𝛿𝑗𝑚: EV vs. GV, 2015–2024 (a) Electric vehicles (b) Gasoline vehicles Notes: Line: sales-weighted mean; dark band: IQR (p25–p75); light band: p10–p90. Selected notable models are annotated [PITH_FULL_IMAGE:figures/full_fig_p016_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Year-specific mean-utility deviations: EV vs. GV [PITH_FULL_IMAGE:figures/full_fig_p017_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Own-price elasticities and Lerner indices [PITH_FULL_IMAGE:figures/full_fig_p023_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Order sensitivity of sequential decompositions [PITH_FULL_IMAGE:figures/full_fig_p026_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Shapley waterfall: 2015–2024 EV-share change [PITH_FULL_IMAGE:figures/full_fig_p028_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Tier-specific Shapley decomposition of the 2015–2024 EV-share change [PITH_FULL_IMAGE:figures/full_fig_p030_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Cumulative Shapley decomposition from the 2015 baseline [PITH_FULL_IMAGE:figures/full_fig_p031_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: No-subsidy counterfactual: dynamic simulation [PITH_FULL_IMAGE:figures/full_fig_p034_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Cross-tier EV diffusion lag, 2015–2024 Notes: EV share is BEV + PHEV + REEV as a fraction of new-vehicle sales in each tier. Tier 1 covers Beijing, Shanghai, Guangzhou, and Shenzhen; New Tier 1 covers 15 new first-tier cities; Tier 2 covers 33 remaining second-tier cities in the panel; Rest covers the 27 small-city residual cells. The double-headed arrow marks the 4-year lag between Tier 1 and Rest crossin… view at source ↗
Figure 12
Figure 12. Figure 12: EV subsidy incidence by city income tier [PITH_FULL_IMAGE:figures/full_fig_p037_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: BEV driving-range trajectories by product, 2015–2024. [PITH_FULL_IMAGE:figures/full_fig_p046_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Distribution of recovered marginal costs, EV vs. GV [PITH_FULL_IMAGE:figures/full_fig_p047_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: Distribution of unobserved marginal-cost residual [PITH_FULL_IMAGE:figures/full_fig_p047_15.png] view at source ↗
read the original abstract

China's electric-vehicle (EV) sales share rose from about 1% in 2015 to roughly 45% in 2024. We evaluate this technology transition with an equilibrium differentiated-products model of the Chinese auto market, and quantify both its attribution and its welfare and reallocation consequences. Every yuan of 2024 EV subsidy delivered about 3.38 yuan of private surplus, but this surplus accrued asymmetrically. Per-capita consumer-surplus loss from subsidy removal is about five times larger in Tier 1 than in the Rest tier; about half of the aggregate welfare loss operates through indirect Wright's-law learning rather than the direct cash transfer; and EV-native firms (BYD, Tesla, New Forces) retain 16-27% of their 2024 EV business under subsidy removal while traditional state-owned manufacturers retain only 11%. A Shapley decomposition into six channels -- Quality, Variety, Battery, Subsidy, Residual, and Market -- attributes the historical 2015-2024 rise primarily to product-quality gains (+45.49%), choice-set expansion (+14.81%), and battery-cost decline (+8.20%). The Subsidy block is negative (-13.63%) because direct purchase subsidies were phased down, not because subsidies reduce demand: a separate counterfactual that removes the 2024 subsidy entirely lowers EV share by 23-33%.

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

3 major / 2 minor

Summary. The manuscript develops an equilibrium differentiated-products model of the Chinese automobile market to analyze the rise in EV sales share from ~1% in 2015 to ~45% in 2024. It employs a Shapley decomposition across six channels (Quality, Variety, Battery, Subsidy, Residual, Market) to attribute the increase primarily to product-quality gains (+45.49%), choice-set expansion (+14.81%), and battery-cost decline (+8.20%). The analysis quantifies welfare consequences of EV subsidies, reporting that each yuan of 2024 subsidy generates 3.38 yuan of private surplus (with ~half the loss from removal operating through indirect Wright's-law learning effects), heterogeneous consumer-surplus impacts across tiers, and differential firm retention rates under subsidy removal.

Significance. If the identification and functional-form assumptions hold, the paper contributes a detailed structural quantification of technology-transition channels and industrial-policy incidence in a major market. The integration of supply-side learning-by-doing with demand-side variety and quality effects, combined with counterfactuals on subsidy removal and firm reallocation, provides concrete policy-relevant numbers. The Shapley decomposition and welfare-multiplier calculations are strengths when supported by robust identification.

major comments (3)
  1. [Counterfactual welfare analysis] Abstract and counterfactual welfare section: the claim that 'about half of the aggregate welfare loss operates through indirect Wright's-law learning' is load-bearing for the headline policy conclusion. This attribution requires that the learning rate on cumulative production is identified separately from demand shifters, unobserved quality trends, and pricing conduct. The manuscript should supply explicit identification arguments, instruments, or robustness checks (e.g., alternative cost specifications) demonstrating that the 50% split is not driven by the chosen normalizations or equilibrium feedback.
  2. [Shapley decomposition] Shapley decomposition (abstract and results section): the reported percentages (Quality +45.49%, Variety +14.81%, Battery +8.20%, Subsidy -13.63%) are computed internally from the same estimated parameters. Without external validation or checks against alternative demand systems, the decomposition risks reflecting functional-form choices rather than true marginal contributions. The paper should report sensitivity of the ordering and magnitudes to perturbations in the demand or cost specification.
  3. [Model and estimation] Model section on demand, supply, and cost functions: the welfare multiplier of 3.38 and the heterogeneous incidence results rest on correct specification and identification of the demand system, oligopolistic pricing, and Wright's-law cost function. The manuscript should clarify how the learning parameter is disentangled from consumer preference heterogeneity and firm-specific quality shocks to ensure the reported surplus and reallocation effects recover structural primitives.
minor comments (2)
  1. [Abstract] Abstract: the precise numerical claims (3.38 multiplier, 23-33% share drop, retention rates 11-27%) should be cross-referenced to the corresponding tables or figures in the main text for immediate traceability.
  2. [Model section] Notation: ensure consistent definition of the six Shapley channels (Quality, Variety, Battery, Subsidy, Residual, Market) when first introduced in the model section.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the detailed and constructive report. We address each major comment below with clarifications on identification and agree to incorporate additional robustness material in the revision.

read point-by-point responses

  1. Referee: [Counterfactual welfare analysis] Abstract and counterfactual welfare section: the claim that 'about half of the aggregate welfare loss operates through indirect Wright's-law learning' is load-bearing for the headline policy conclusion. This attribution requires that the learning rate on cumulative production is identified separately from demand shifters, unobserved quality trends, and pricing conduct. The manuscript should supply explicit identification arguments, instruments, or robustness checks (e.g., alternative cost specifications) demonstrating that the 50% split is not driven by the chosen normalizations or equilibrium feedback.

    Authors: The learning rate is identified from the supply-side cost equation using time-series and cross-firm variation in cumulative production, which is driven by lagged demand and policy realizations that are orthogonal to current unobserved demand shocks after conditioning on the demand-side instruments and fixed effects. Demand shifters and quality trends are absorbed in the random-coefficients demand system estimated with BLP-style instruments; pricing conduct enters only through the equilibrium markup term. We will add an explicit identification subsection in the model section and report robustness to alternative cost specifications (log-linear versus power-law learning) and to shutting down equilibrium feedback in the counterfactuals. These checks confirm that the approximate 50% indirect-learning share is not an artifact of the baseline normalizations. revision: yes

  2. Referee: [Shapley decomposition] Shapley decomposition (abstract and results section): the reported percentages (Quality +45.49%, Variety +14.81%, Battery +8.20%, Subsidy -13.63%) are computed internally from the same estimated parameters. Without external validation or checks against alternative demand systems, the decomposition risks reflecting functional-form choices rather than true marginal contributions. The paper should report sensitivity of the ordering and magnitudes to perturbations in the demand or cost specification.

    Authors: We agree that the Shapley values are model-internal. In the revision we will add a dedicated sensitivity table that recomputes the full decomposition after (i) scaling the random-coefficient standard deviations by ±20% around their point estimates, (ii) replacing the nested-logit outside-good share with a simple logit, and (iii) altering the Wright's-law exponent within its 95% confidence interval. The ordering and the dominance of the Quality and Variety channels remain stable under these perturbations. revision: yes

  3. Referee: [Model and estimation] Model section on demand, supply, and cost functions: the welfare multiplier of 3.38 and the heterogeneous incidence results rest on correct specification and identification of the demand system, oligopolistic pricing, and Wright's-law cost function. The manuscript should clarify how the learning parameter is disentangled from consumer preference heterogeneity and firm-specific quality shocks to ensure the reported surplus and reallocation effects recover structural primitives.

    Authors: Consumer preference heterogeneity is identified from cross-sectional choice variation conditional on observed characteristics and prices (via the random-coefficient distribution), while the learning parameter enters only the marginal-cost function as a function of cumulative output. Firm-specific quality shocks are recovered as the residual in the demand inversion after controlling for observed attributes and are therefore orthogonal to the cost-side learning term. We will expand the estimation subsection to spell out this separation and to note that the welfare and reallocation counterfactuals are computed from the joint equilibrium of the estimated primitives. revision: yes

Circularity Check

0 steps flagged

No significant circularity in derivation chain

full rationale

The paper estimates a structural differentiated-products equilibrium model on observed Chinese auto market data, then runs counterfactuals to compute welfare multipliers and applies Shapley decomposition to attribute the EV share rise across six channels. These operations use fitted parameters to simulate alternative equilibria and marginal contributions; they do not reduce by construction to the inputs via self-definition, renaming of fitted values as predictions, or load-bearing self-citations. The claims rest on external data identification and standard structural methods rather than tautological equivalence.

Axiom & Free-Parameter Ledger

2 free parameters · 2 axioms · 0 invented entities

Because only the abstract is available, the ledger is necessarily incomplete. The model must contain numerous estimated demand and cost parameters plus functional-form assumptions for the learning curve and equilibrium conditions that are not visible here.

free parameters (2)
  • Demand and cost parameters
    Standard in differentiated-products models; exact count and values unknown from abstract.
  • Wright's-law learning rate
    Fitted parameter governing how cumulative production reduces future costs; central to the indirect welfare channel.
axioms (2)
  • domain assumption Market clears in a differentiated-products Nash-Bertrand equilibrium each period
    Invoked to close the model and generate counterfactual prices and quantities.
  • domain assumption Cost reductions follow a deterministic Wright's-law function of cumulative output
    Required to separate direct subsidy effects from indirect learning effects.

pith-pipeline@v0.9.1-grok · 5787 in / 1407 out tokens · 41545 ms · 2026-06-29T02:16:19.877792+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

34 extracted references

  1. [1]

    Econometrica , volume =

    Berry, Steven and Levinsohn, James and Pakes, Ariel , title =. Econometrica , volume =

  2. [2]

    Journal of Political Economy , volume =

    Berry, Steven and Levinsohn, James and Pakes, Ariel , title =. Journal of Political Economy , volume =

  3. [3]

    and Haile, Philip A

    Berry, Steven T. and Haile, Philip A. , title =. Econometrica , volume =

  4. [4]

    Econometrica , volume =

    Nevo, Aviv , title =. Econometrica , volume =

  5. [5]

    Journal of Political Economy , volume =

    Gowrisankaran, Gautam and Rysman, Marc , title =. Journal of Political Economy , volume =

  6. [6]

    Journal of Political Economy , volume =

    Petrin, Amil , title =. Journal of Political Economy , volume =

  7. [7]

    American Economic Review , volume =

    Fan, Ying , title =. American Economic Review , volume =

  8. [8]

    Econometrica , volume =

    Sweeting, Andrew , title =. Econometrica , volume =

  9. [9]

    Journal of the Association of Environmental and Resource Economists , volume =

    Li, Shanjun and Tong, Lang and Xing, Jianwei and Zhou, Yiyi , title =. Journal of the Association of Environmental and Resource Economists , volume =

  10. [10]

    , title =

    Ryan, Stephen P. , title =. Econometrica , volume =

  11. [11]

    , title =

    Fowlie, Meredith and Reguant, Mar and Ryan, Stephen P. , title =. Journal of Political Economy , volume =

  12. [12]

    American Economic Journal: Economic Policy , volume =

    Springel, Katalin , title =. American Economic Journal: Economic Policy , volume =

  13. [13]

    American Economic Review , volume =

    Acemoglu, Daron and Aghion, Philippe and Bursztyn, Leonardo and Hemous, David , title =. American Economic Review , volume =

  14. [14]

    and Jaffe, Adam B

    Newell, Richard G. and Jaffe, Adam B. and Stavins, Robert N. , title =. Quarterly Journal of Economics , volume =

  15. [15]

    American Economic Review , volume =

    Popp, David , title =. American Economic Review , volume =

  16. [16]

    Journal of Economic Perspectives , volume =

    Thompson, Peter , title =. Journal of Economic Perspectives , volume =

  17. [17]

    Lanier , title =

    Benkard, C. Lanier , title =. Review of Economic Studies , volume =

  18. [18]

    American Economic Review , volume =

    Bollinger, Bryan and Gillingham, Kenneth , title =. American Economic Review , volume =

  19. [19]

    , title =

    Shorrocks, Anthony F. , title =. Income Inequality: Economic Disparities and the Middle Class in Affluent Countries , publisher =. 2013 , pages =

    2013

  20. [20]

    , title =

    Shapley, Lloyd S. , title =. Contributions to the Theory of Games II , publisher =. 1953 , pages =

    1953

  21. [21]

    , title =

    Wollmann, Thomas G. , title =. American Economic Review , volume =

  22. [22]

    International Journal of Industrial Organization , volume =

    Hao, Yiran and Hao, Yu , title =. International Journal of Industrial Organization , volume =

  23. [23]

    Environmental science & technology , volume=

    Compliance by design: Influence of acceleration trade-offs on co2 emissions and costs of fuel economy and greenhouse gas regulations , author=. Environmental science & technology , volume=. 2017 , publisher=

    2017

  24. [24]

    The review of economic studies , volume=

    The concept of monopoly and the measurement of monopoly power , author=. The review of economic studies , volume=. 1934 , publisher=

    1934

  25. [25]

    and Mansur, Erin T

    Holland, Stephen P. and Mansur, Erin T. and Muller, Nicholas Z. and Yates, Andrew J. , title =. American Economic Review , volume =

  26. [26]

    , title =

    Sallee, James M. , title =. American Economic Journal: Economic Policy , volume =

  27. [27]

    , title =

    Muehlegger, Erich and Rapson, David S. , title =. Journal of Public Economics , volume =

  28. [28]

    Journal of Environmental Economics and Management , volume =

    Xing, Jianwei and Leard, Benjamin and Li, Shanjun , title =. Journal of Environmental Economics and Management , volume =

  29. [29]

    Rapidly Falling Costs of Battery Packs for Electric Vehicles , journal =

    Nykvist, Bj\"orn and Nilsson, M. Rapidly Falling Costs of Battery Packs for Electric Vehicles , journal =

  30. [30]

    and Trancik, Jessika E

    Ziegler, Micah S. and Trancik, Jessika E. , title =. Energy & Environmental Science , volume =

  31. [31]

    American Economic Journal: Macroeconomics , volume =

    Aghion, Philippe and Cai, Jing and Dewatripont, Mathias and Du, Luosha and Harrison, Ann and Legros, Patrick , title =. American Economic Journal: Macroeconomics , volume =

  32. [32]

    Wright, T. P. , title =. Journal of the Aeronautical Sciences , volume =

  33. [33]

    Review of Economic Studies , volume =

    Reynaert, Mathias , title =. Review of Economic Studies , volume =

  34. [34]

    and Mansur, Erin T

    Holland, Stephen P. and Mansur, Erin T. and Muller, Nicholas Z. and Yates, Andrew J. , title =. Journal of the Association of Environmental and Resource Economists , volume =