Discussant Comments

Wang et al. – ‘Dragon’s Roar: Pollution Abatement Costs in Chinese Manufacturing’

Juan Vélez-Velásquez

2025-04-01

Key Contribution

• Estimates firm-level marginal abatement costs (MAC) of SO₂
  
• Uses a by-production framework with GMM
  
• Combines structural modeling with Chinese manufacturing data
  
• A novel and timely methodological contribution

• Make substantive contribution to empirical environmental economics.
  
• not an expert on the topic but I do not think you have to be one to appreciate the value of having better estimates of abatement costs.
• They use a by-production structure and GMM to estimate the shadow price of SO₂ abatement across Chinese firms.
  

Focus of My Comments

1. Market structure assumptions
   
2. Suggestions for future work
   
3. Modeling and identification choices

I’ll focus my remarks on three areas:
(1) the assumptions about price-taking behavior,
(2) suggestions for extending the framework,
and (3) modeling and identification comments

Market Structure Assumptions

• Assumes competitive prices for both inputs and outputs
  
• Likely valid for inputs
  
• More questionable for output prices

The paper assumes competitive pricing in both input and output markets.
While this is plausible for inputs, output prices are more complex.
Some firms may operate in differentiated markets or have pricing power.
That could bias MAC estimates if firms are passing pollution control costs to consumers.

Why This Matters

• Output price not exogenous if firm has market power
  
• Abatement may reflect strategic pricing, not only cost minimization

Recommendation:
Acknowledge this assumption and discuss how estimates might change under imperfect competition

This matters because if output prices are not exogenous, observed firm behavior might reflect pricing strategy—not just technical efficiency.
Even a short discussion of this would help strengthen the credibility and transparency of the estimates. Price taking has been a “feature” of production function estimation.

Suggestions for Future Work – Oligopoly Instruments

• Ackerberg & De Loecker (2024):
  • “Oligopoly instruments”: use competitor productivity or fixed factors
    
  • Improve identification when firms have market power

Highly relevant in mixed-ownership settings like China

For future research, the framework from Ackerberg and De Loecker offers a promising path.
By using variation in competitor behavior or productivity, we can better identify production functions under imperfect competition.
This could be particularly valuable in the Chinese context, where competition structures are often opaque or politicized.

Modeling & Identification Choices

Materials vs. Coal

• Good output: materials input (M)
  
• Emissions: coal input (C) directly
  
• Appears inconsistent — but:

Footnote: coal is embedded in (M) due to missing price data
Modeled separately for emissions using observed physical quantities

There’s an apparent inconsistency in how coal is treated across the two production functions.
Coal is embedded in materials in the good-output function, but used directly in the emissions function.
This is a data-driven modeling choice, and it’s explained in a footnote. I think that explanation deserves a place in the main text.

Identification Strategy

• Based on moment conditions: \((E[\text{instruments} \cdot \text{residuals}] = 0)\)
• This is standard and valid within GMM
  
• But it’s very mechanical — may not highlight the economic intuition

Suggestion:
Complement the GMM moment conditions with an explanation of the source of exogenous variation
(e.g., what variation is identifying the elasticities? firm heterogeneity? industry shocks?)

The identification argument is valid — it follows GMM logic faithfully.
But I found it very mechanical: essentially saying “this expected value is zero.”
I encourage the authors to also describe the source of identifying variation in economic terms.
This might help reach a broader audience and enhance interpretability.

Identification Strategy (II)

Both identification arguments are analogous

• Proxy for good productivity - some orthogonality
• Proxy for bad productivity - some orthogonality

Suggestion:
Use this symmetry to reduce the size and burden of that section

I think the identification section can be shorter and equally effective making use of the symmetry in the identification arguments.

Final Thoughts

• Strong and well-motivated paper
  
• Comments aimed to clarify assumptions and inspire extensions
  
• Looking forward to future iterations

To wrap up: this is a strong paper with a lot of potential.
My comments are offered in the spirit of strengthening clarity and opening new avenues for expansion.
I look forward to the conversation. Thank you.