Poison or catalyst? How do energy saving targets matter for firm-level productivity in China

The study addresses how mandatory energy saving targets under China’s Top-10000 Enterprises Energy Conservation and Low Carbon Program influence firms’ total factor productivity (TFP). Against the backdrop of China’s substantial share of global CO2 emissions and reliance on energy-intensive industries, the government assigns binding energy conservation targets to major energy-consuming enterprises to reduce emissions. While TFP is central to gauging growth quality, prior research offers mixed evidence on how environmental target-setting affects firm productivity and often overlooks the role of target intensity. The paper poses three questions: (1) Do energy saving targets affect corporate TFP? (2) What is the overall nature (functional form) of this effect? (3) Through which channels, and with what heterogeneity across regions, ownership types, and industries, do targets influence TFP? The authors hypothesize an inverted-U-shaped relationship between target intensity and TFP (Hypothesis 1) and propose that market share expansion mediates the effect of targets on TFP (Hypothesis 2).

China’s environmental governance involves target-based management at both government and enterprise levels. Extensive literature examines provincial energy intensity and emission reduction targets, finding they can reduce emissions and influence outcomes like wind energy development, PM2.5 control, pollutant emissions, and economic goals. At the enterprise level, studies on the Top-1000 and Top-10000 programs discuss target allocation, implementation weaknesses, and performance reporting, with state-owned and centrally affiliated firms often receiving higher targets. Evidence on performance impacts includes decreased profitability and increased costs under Top-1000, energy savings via efficiency and scale changes under Top-10000, and effects on exports and employment. Only two English-language studies assess TFP impacts: Ai et al. (2021) found negative TFP effects in chemical firms under Top-1000; Filippini et al. (2020) found positive effects for iron and steel firms. Gaps include limited attention to enterprise-level targets versus region-level targets, under-examination of the Top-10000 program relative to Top-1000, lack of analysis on channels (e.g., market share), and limited exploration of heterogeneity. The authors propose that compliance costs versus benefits can produce either productivity-enhancing innovation (Porter hypothesis) or crowding out of resources, implying a potential inverted-U relationship; market share expansion is posited as a mediating channel linking targets to TFP.

Data: The authors combine (1) the 2011 NDRC Top-10000 program list providing firm-specific absolute energy saving targets (tce) for 16,373 organizations (including 14,249 industrial enterprises) and (2) the China Industrial Enterprise Database (CIED) for 2012 and 2013 (311,557 observations in 2012 and 345,101 in 2013). After excluding non-industrial organizations (schools, transportation, trading, hotels/restaurants; N=2,124), they match the remaining 14,249 industrial enterprises to CIED, yielding 12,299 firm-year matches, and apply data screening (missing key variables; accounting inconsistencies; age ≤1 year; employees <10) resulting in a final two-year panel of 10,667 firms (21,334 observations).

Variables: Dependent variable is firm-level TFP. TFP is estimated via a Cobb-Douglas production function with inputs labor (employees), capital (net fixed assets), and intermediate inputs, applying the Levinsohn-Petrin (LP) method to address zero investment; Olley-Pakes (OP) and fixed-effects (FE) TFP estimates are used in robustness checks. Output is proxied by total output; intermediate inputs equal COGS + selling + administrative + financial expenses − current depreciation − payroll payable. The TFP measure lnA_it is computed as residual from the estimated production function.

Key independent variable: target intensity, constructed as the enterprise’s absolute energy-saving target (Top-10000 list; fixed over 12th FYP) divided by firm size (natural logarithm of total assets) to avoid confounding from absolute target levels and to introduce cross-sectional variation over 2012–2013. A quadratic term (target^2) captures potential non-linearity.

Mediator: market share, measured as the firm’s value-added share within its region-industry-year market (percentage of firm added value over the region-industry total added value).

Controls: profitability (roa = net profit/fixed assets), leverage (lev = debt/assets), fixed assets ratio (fixed = fixed assets/total assets, logged), size (ln total assets), SOE ownership dummy, export status dummy (export delivery value >0), and age (current year − start year + 1). Year, province, and industry fixed effects are included; additional specifications control for year-industry, year-province, and industry-province fixed effects.

Models: Step 1 estimates TFP on target and target^2 with controls and fixed effects to test for linear and non-linear effects. Step 2 applies mediation analysis (Baron and Kenny framework): (i) regress market share on target and target^2; (ii) regress TFP on market share; (iii) regress TFP on target, target^2, and market share. Robustness checks include: winsorizing target at 0.5%/99.5%; alternative TFP measures (FE, OP); controlling contemporaneous policies (Low-Carbon City Pilots; Carbon Emissions Trading System Pilots); excluding four municipalities (Beijing, Shanghai, Tianjin, Chongqing); additional interaction fixed effects; and causal mediation analysis (Imai et al., 2010) to quantify mediated effects and address potential selection bias.

Sample statistics: 21,334 observations; mean TFP (LP) ≈ 5.614 (SD 1.291); mean target ≈ 2.327 (SD 5.791); market share mean ≈ 0.927 (SD 2.575). Variance inflation factors are below 3.42, indicating no severe multicollinearity.

• Baseline linear effect: Target intensity is positively associated with TFP (Model 1: β_target = 0.0677, p<0.01), indicating productivity gains at lower target levels.
• Non-linear effect (inverted-U): Including target^2 yields significantly negative coefficients across specifications. In the fully controlled model (Model 6): β_target = 0.01429 (p<0.01), β_target^2 = −0.00019 (p<0.01), with a turning point at approximately 37.2. Thus, TFP rises with target intensity up to ≈37.2 and declines thereafter. The average target intensity (≈2.327) is far below the turning point; the maximum (≈178.9) exceeds it, implying that very high targets can hinder TFP.
• Control variables: Larger size, higher profitability, and exporting status positively relate to TFP; higher leverage, a higher fixed asset ratio, SOE status, and greater age are negatively associated with TFP.
• Mediation by market share: Targets increase market share nonlinearly (Model 8: β_target = 0.1690, p<0.01; β_target^2 = −0.0002, p<0.01). Market share positively affects TFP (Model 10: β = 0.1612, p<0.01). In the joint model (Model 12), effects remain: β_target = 0.0143 (p<0.01), β_target^2 = −0.0002 (p<0.01), β_marketshare = 0.1610 (p<0.01); R^2 ≈ 0.7289. Sobel test p<0.001; bootstrap (1000) indirect effect Coef ≈ 0.0195, Z ≈ 17.94, 95% CI [0.017375, 0.0216], supporting mediation.
• Robustness: Results persist after winsorizing target (Models 13–15); using FE and OP TFP measures (Models 16–20); controlling LCCP and CETSP policy dummies (Models 21–25); excluding four municipalities (Models 26–28); and adding year-industry, year-province, and industry-province fixed effects (Models 29–37). Causal Mediation Analysis indicates about 72.38% of the total effect is mediated by market share (ACME ≈ 0.0325; direct effect ≈ 0.0123; total effect ≈ 0.0447).
• Heterogeneity: Inverted-U holds across subgroups, but turning points differ: Eastern region turning point ≈ 19.6; non-eastern ≈ 49.2. Non-SOEs turning point ≈ 30.8; SOEs ≈ 69. By industry: mining ≈ 41.2; manufacturing ≈ 40.6; power, gas, and water supply ≈ 26.6. These differences suggest varying capacity and incentive structures shape how target intensity translates into productivity.

The findings reconcile prior contradictory evidence by demonstrating a non-linear (inverted-U) relationship between energy saving target intensity and firm TFP. Modest target pressure appears to stimulate productivity-enhancing adjustments—such as process improvements and efficient resource reallocation—consistent with a Porter-type effect. As target intensity rises beyond a threshold, compliance costs crowd out productivity-enhancing investments (e.g., R&D, human capital), diminishing TFP. The identified mediation through market share indicates that targets contribute to competitive repositioning: firms that effectively comply improve market share, and this scale and market power effect bolsters productivity. The substantial mediated proportion suggests that market dynamics are a central channel linking environmental regulation to operational efficiency. Heterogeneous turning points across regions, ownership types, and industries underscore the importance of context: more competitive or technologically advanced environments (eastern, non-SOEs, some industries) may exhaust productivity gains at lower target intensities, while entities with greater access to resources or slack (SOEs, some non-eastern firms) can sustain benefits under higher targets before facing diminishing returns. These insights highlight the policy relevance of tailoring target intensities to local industrial conditions and ownership structures to maximize productivity and environmental gains.

This study compiles a large-scale, enterprise-level dataset linking Top-10000 energy saving targets with firm outcomes and constructs an intensity-based measure of target pressure to examine effects on TFP. It establishes an inverted-U-shaped relationship between target intensity and TFP, identifies market share as a key mediating mechanism, and documents significant heterogeneity across regions, ownership, and industries. The results suggest that appropriately calibrated targets can simultaneously improve environmental outcomes and productivity, while overly stringent targets risk undermining firm performance. Policy recommendations include: calibrating target intensity relative to firms’ characteristics and sectoral contexts; setting comparatively higher targets for non-eastern firms, SOEs, and manufacturing firms to leverage their greater potential; and using market share evolution as an indicator to guide and adjust targets. Future research should expand the time horizon to capture long-run effects and incorporate technological innovation metrics to better understand dynamic productivity responses to environmental target-setting.

• Temporal limitation: Analysis is restricted to 2012–2013 due to changes in CIED indicators (notably missing net profit and selling expenses in 2014), preventing examination of longer-term effects and dynamic adjustment processes.
• Mechanism scope: The study primarily examines resource allocation and market share as the mediating mechanism; it does not directly analyze technological innovation or improvement channels due to data limitations.
• Fixed target values: Although target intensity is scaled by firm size to introduce variation, the underlying absolute targets are fixed for the 12th FYP, which may limit identification of within-firm temporal variation beyond cross-sectional differences over two years.