Using China's Annual Survey of Industrial Firms (ASIF) for 2001–2007, Howell (2017) estimated a Crépon–Duguet–Mairesse (CDM)-type structural innovation model with matched panels and found that production subsidies modestly raise measured innovation sales intensity but are associated with small or negative effects on R&D participation and with negative coefficients on firm total factor productivity, casting doubt on China's 'picking winners' subsidy strategy. The paper's authors acknowledge three unresolved limitations that jointly define an open empirical problem: (1) the 2001–2007 window captures only shorter-term effects, leaving long-horizon impacts on risky innovation projects, eventual productivity, and diffusion/spillovers unobserved; (2) subsidy allocation is endogenous, and matching plus Mundlak-type corrections cannot rule out unobserved time-varying selection in the absence of quasi-experimental variation; (3) the innovation output measure (new product/process sales) conflates in-house knowledge creation with adoption or purchase of external technologies, so it is unknown which channel subsidies actually drive and how each maps into productivity. Later work has addressed pieces of this question with conflicting results — a structural endogenous-treatment estimate finds positive productivity effects in machinery (1998–2007), while a listed-firm study over 2007–2018 finds subsidies negatively associated with ex-post productivity growth — but no study provides a credibly identified, long-horizon evaluation of production subsidies on the manufacturing firm universe with the innovation-output channel decomposed. The problem — determining the causal effect of production subsidies on indigenous innovation and firm productivity in Chinese manufacturing over longer horizons, with credible identification of endogenous subsidy allocation and a decomposition of subsidy-induced innovation output into internal knowledge creation versus external technology acquisition — remains open.

Background

China's rapid industrial upgrading has been accompanied by large-scale state intervention, including direct 'production subsidies' (current-account transfers to firms, as distinct from targeted R&D grants) allocated by governments at multiple levels in the hope of 'picking winners' — selecting recipient firms that will become successful indigenous innovators with large welfare payoffs. Whether this strategy works is a central question in the economics of industrial policy and state capitalism, and firm-level microdata are the main evidence base.

The key dataset is the Annual Survey of Industrial Firms (ASIF), China's official survey of all state-owned and above-scale (sales above 5 million RMB) non-state manufacturing firms, which records, among other things, subsidy receipts and the share of revenue from 'new products' — goods new to the firm or to the market, the standard firm-level innovation-output measure in this literature. Prior work had established that 'indigenous innovation' (innovation generated by a firm's own knowledge creation rather than absorption of foreign technology) is heterogeneous in risk and matters for new-firm survival in China (Howell 2015). Building on this, Howell (2017) studied majority-private manufacturing firms in the ASIF 2001–2007 panel, measuring subsidies as the share of firm revenue received as production subsidies, and estimating a CDM-type structural model (the Crépon–Duguet–Mairesse system, which links an R&D-input equation to an innovation-output equation and then to a productivity equation) combined with propensity-score matching on pre-treatment means and panel random-effects estimators with Mundlak corrections (device that proxies unobserved firm effects by within-firm means of regressors, mitigating — but not eliminating — correlated unobserved heterogeneity). The findings: production subsidies modestly raise innovation sales intensity (coefficients around 0.080–0.199 across technology classes), have small or negative associations with R&D participation and R&D intensity, and are associated with negative firm total factor productivity (TFP; the residual output not explained by measured inputs) coefficients (−0.073 to −0.019); state-backed 'successful innovators' also show smaller TFP gains than non-backed ones. The authors conclude the evidence does not support the expectation that the social payoffs of picking winners reliably outweigh the efficiency losses.

The paper itself acknowledges, however, that this verdict rests on a short window, a partially treated identification problem, and an output measure that cannot distinguish innovation modes. First, the ASIF 2001–2007 window is too short to observe effects on risky, long-horizon innovation projects and their eventual productivity, diffusion, and spillover consequences. Second, subsidies are not randomly assigned: firms are plausibly selected on unobserved, time-varying characteristics (political connections, policy expectations), and no quasi-experimental variation in production-subsidy allocation is exploited, so even the matched-panel estimates are not fully causal. Third, new product sales aggregate revenue from in-house R&D and from adoption or purchase of external technologies (licensing, embodied equipment, technology transfer), so it is unknown whether subsidy-driven measured innovation reflects internal knowledge creation or technology acquisition, and how each channel differentially affects productivity. Sun and Du (2010), using China's 2004 economic census, had already shown that in-house R&D and external technology purchase are distinct and unequally productive innovation channels, underscoring that the aggregation matters.

Problem Statement

Determine the causal effect of government production subsidies on indigenous innovation and firm productivity in Chinese manufacturing, over horizons extending beyond the ASIF 2001–2007 window, resolving the three linked limitations acknowledged by Howell (2017): (1) estimate the longer-run effects of subsidy receipt on risky, long-horizon innovation output, on eventual firm productivity, and on diffusion/spillovers to other firms, rather than only the shorter-term matched-panel effects; (2) identify those effects credibly despite the endogenous, non-random allocation of production subsidies — addressing unobserved, time-varying selection into subsidy receipt that matching and Mundlak-type corrections cannot rule out; and (3) decompose the subsidy-induced increase in measured innovation output (new product/process sales) into internal knowledge creation versus adoption or purchase of external technologies, establishing how much of the measured effect each channel accounts for and how each channel differentially affects productivity. These three components jointly answer one question — whether China's production-subsidy 'picking winners' strategy causes durable gains in genuinely indigenous innovation and productivity.

References

  1. Howell, A. (2017). Picking 'winners' in China: Do subsidies matter for indigenous innovation and firm productivity? China Economic Review 44:154–165. DOI: 10.1016/j.chieco.2017.04.005, https://doi.org/10.1016/j.chieco.2017.04.005
  2. Howell, A. (2015). 'Indigenous' innovation with heterogeneous risk and new firm survival in a transitioning Chinese economy. Research Policy 44(10):1866–1876. DOI: 10.1016/j.respol.2015.06.012, https://doi.org/10.1016/j.respol.2015.06.012
  3. Sun, Y. and Du, D. (2010). Determinants of industrial innovation in China: Evidence from its recent economic census. Technovation 30(9–10):540–550. DOI: 10.1016/j.technovation.2010.05.003, https://doi.org/10.1016/j.technovation.2010.05.003
  4. Li, M., Jin, M. and Kumbhakar, S.C. (2022). Do subsidies increase firm productivity? Evidence from Chinese manufacturing enterprises. European Journal of Operational Research 303(1):388–400. DOI: 10.1016/j.ejor.2022.02.029, https://doi.org/10.1016/j.ejor.2022.02.029
  5. Branstetter, L., Li, G. and Ren, M. (2023). Picking winners? Government subsidies and firm productivity in China. Journal of Comparative Economics 51(4):1186–1199. DOI: 10.1016/j.jce.2023.06.004, https://doi.org/10.1016/j.jce.2023.06.004
  6. Yang, C.-H. (2022). External technological sources, subsidies and firm productivity in China. Applied Economics 55(47):5572–5590. DOI: 10.1080/00036846.2022.2140113, https://doi.org/10.1080/00036846.2022.2140113
  7. Boeing, P., Eberle, J. and Howell, A. (2022). The impact of China's R&D subsidies on R&D investment, technological upgrading and economic growth. Technological Forecasting and Social Change 174:121212. DOI: 10.1016/j.techfore.2021.121212, https://doi.org/10.1016/j.techfore.2021.121212
  8. Dai, X. and Wang, F. (2019). Does the high- and new-technology enterprise program promote innovative performance? Evidence from Chinese firms. China Economic Review 57:101330. DOI: 10.1016/j.chieco.2019.101330, https://doi.org/10.1016/j.chieco.2019.101330
  9. Bulman, D.J., Yan, X. and Zhang, Q. (2022). Picking Losers: How Career Incentives Undermine Industrial Policy in Chinese Cities. Journal of Development Studies 58(6):1102–1123. DOI: 10.1080/00220388.2022.2043280, https://doi.org/10.1080/00220388.2022.2043280
  10. Shi, J., Sadowski, B.M., Zeng, X., Dou, S., Xiong, J., Song, Q. and Li, S. (2023). Picking winners in strategic emerging industries using government subsidies in China: the role of market power. Humanities and Social Sciences Communications 10:article 394. DOI: 10.1057/s41599-023-01910-9, https://doi.org/10.1057/s41599-023-01910-9
  11. Yan, X., Yu, S. and Zheng, X. (2025). Is government subsidy efficient? Evidence from SMEs' automation in China. China Economic Review 94:102566. DOI: 10.1016/j.chieco.2025.102566, https://doi.org/10.1016/j.chieco.2025.102566