## Appendix I. (wpiea2021228-print-pdf)

## Source details

**Canonical URL:** [Appendix I. (wpiea2021228-print-pdf)](https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021228-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2021/english/wpiea2021228-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2021/english/wpiea2021228-print-pdf.pdf.json)

---

### I. Introduction and Context
- Transition to low-emission and sustainable development requires a massive reallocation of capital and adoption of clean production technology, likely generating significant and systematic impacts on financial stability and macroeconomic conditions.
- Transition risks are a new category and remain relatively underexplored compared with physical risks.
- The paper investigates how an environmental policy aiming at emission abatement affects financial stability from theoretical and empirical perspectives, focusing on channels through which the transition impacts a commercial bank’s balance sheet.

### Theoretical framework: Environmental DSGE (E-DSGE) model
- Purpose and mechanism:
  - Construct an environmental dynamic stochastic general equilibrium (E-DSGE) model to illustrate reactions of productive firms and financial institutions to policy regulations (e.g., emission caps).
  - Predicts a financial accelerator mechanism whereby tightening environmental policies lead entrepreneurs to default endogenously on the outstanding value of their debt.
- Key channels:
  - Environmental policies affect the value of collateral pledged against borrowing.
  - Reduced collateral values increase endogenous default probabilities.
  - This generates an excess premium on loans (difference between contracted loan rates and risk-free rates).
- Core structure and equations (referenced):
  - Entrepreneur profit Π(ω) = ω R_K q_k k_e − R_z b_e (Eq. 3.2).
  - Default threshold from Π(ω)=0 (Eq. 3.3).
  - Emission constraint tightened via environmental policy shock ε_M,t (AR(1)) (Eqs. 3.8–3.9).
  - Bank participation constraint reflecting expected loan returns net of monitoring and seizure (Eq. 3.14).
  - Monetary policy: Taylor-type rule (Eq. 3.25).
- Parameterization (time unit: quarters; Table 1 values preserved):
  - β Households Discount factor: 0.93
  - β_e Entrepreneurs Discount factor: 0.92
  - σ_c Elasticity of substitution for consumption: 1.353
  - ν_L Labor disutility parameter: 1
  - η Labor supply aversion: 2
  - δ_k Capital depreciation parameter: 0.025
  - ψ_k Capital adjustment cost: 6.7
  - α Capital Share: 0.46
  - γ_g Energy Inputs Share: 0.099
  - γ_ng Pollution Inputs Share: 0.09
  - θ Intensity Target coefficient: 0.05
  - F_j(ω̄_g) Probability of default (Green): 0.009
  - F_j(ω̄_ng) Probability of default (Non-Green): 0.010
  - ξ Price Elasticity of Demand for Good n: 6
  - θ Calvo’s Price Parameter for Nominal Rigidities: 0.568
  - ρ_R Monetary Policy Inertia: 0.73
  - ρ_Y Monetary Policy Reaction to Y: 0.109
  - ρ_π Monetary Policy Reaction to π: 1.487
  - β_b Banks Discount factor: 0.93
  - ρ_b Banks Capital ratio: 0.08
  - Θ Cost of deviation from the required capital ratio: 0.25
  - μ_j Monitoring Cost: 0.21
  - σ_g Size of Green Firms: 0.3
  - ρ_M Persistency of Environ. Policy shock: 0.97
  - σ_M Standard deviation on Environ. Policy shock: 0.01
- Calibration notes:
  - Discount factor β = β_b = 0.93 targets average Chinese one-year official lending rate of 7.4 percent over 1991–2012; entrepreneurs face lower β_e = 0.92.
  - γ_j = 0.099 set so that averaged energy expenditures account for 12 percent of GDP.
  - To achieve a 20 percent decrease in emissions, environmental policy shock persistence ρ_M = 0.97 and σ_M = 0.01.
  - Default data: average probability of default set to 0.009 (green) and 0.010 (non-green) from CBIRC corporate credit database.

### Impulse-response scenario and simulated quantitative results (Section IV)
- Scenario:
  - Government tightens environmental regulatory standards so that the pollution constraint decreases by 20 percent (simulation: one standard deviation negative shock to M; impulse responses reported as percentage deviations from steady state over a 20-quarter period).
- Simulated model findings:
  - Productivity of non-green sectors dampened by around 2 percent following the emission cap.
  - Default probabilities:
    - Default rates increase by around 7.5 percent for the non-green sector and 1.8 percent for the green sector (model-simulated percentage increases).
  - Excess premium responses:
    - Excess premium increases by around 0.5 percent for the non-green sector and 0.1 percent for the green sector.
  - Banking sector responses and spillovers:
    - Banks cut the supply of loans due to reduced bank capital (higher defaults and lower asset prices).
    - Banks charge higher lending rates to both non-green and, to a lesser extent, green entrepreneurs to recover monitoring costs and forgone repayments.
    - Green sector experiences negative spillovers through higher borrowing costs and lower demand for external funds; some green entrepreneurs face increased default risk (smaller magnitude).
- Summary conclusion:
  - Environmental policy targeting non-green emissions can spill over to the green sector via banking-sector pricing and balance-sheet channels, producing large negative effects on asset prices and borrowing conditions.

### Empirical strategy and data (Sections V–VI)
- Quasi-natural experiment:
  - Clean Air Action Plan enacted on 12 September 2013; Jiangsu provincial launch 4 January 2014.
  - Clean Air Action Plan set quantitative air quality improvement goals and required local governments to tighten regulation on manufacturing, especially high-emission or high-energy-consumption industries.
- Data:
  - Unique micro-level dataset containing 1.3 million loans granted to nonfinancial firms located in Jiangsu Province.
  - Main sample for analysis: loans granted between 1 September 2012 and 31 December 2014; repayment status traced to 31 March 2016 (in many specifications).
  - Sample in main analysis: 52 commercial banks; 452,208 loans to 60,856 firms (loans in Table 2 sample: 351,888 loans to low-polluting firms; 100,320 loans to high-polluting firms).
  - Loan-level details: unique firm identifier, firm fundamentals (age, size, ownership, location), bank characteristics, loan amount, maturity, credit guarantee, issuing date, delinquency status; monthly mandatory loan updates.
- Prefecture context:
  - Population 80.4 million in 2018; area 102,600 km^2.
  - 2018 GDP US$1.377 trillion (RMB9.2 trillion).
  - 2013 annual industrial SO2 emissions per unit land area: 8.48 tonnes/km^2; CO2: 19.51 tonnes/km^2.
  - Total commercial bank assets in Jiangsu: RMB 16782 billion (around US$2,494 billion), 8 percent of national commercial banking industry as of 2018.
- Outcome variables and empirical design:
  - Outcome variables: loan default indicator (1 if defaulted, 0 otherwise); loan spread measured as percentage deviation of lending rate of each loan from the benchmark rate.
  - Identification: difference-in-differences (DID) using before-and-after policy variation and cross-industry variation given emission-intensity heterogeneity across industries.
  - Time windows:
    - Pre-regulation period: 1 September 2012–10 September 2013.
    - Interim period: 12 September 2013–4 January 2014.
    - Post-regulation period: 6 January 2014–31 December 2014.
  - Main regression (Eq. 5.1): y_l b f t = β0 + β1 Action + β2 Action*Treat + β3 Treat + controls + fixed effects + u_l b f t.
  - Fixed effects: time, industry, bank, prefecture; robustness adds bank*year and prefecture*year.
  - Standard errors clustered at industry level (robustness: cluster by bank and bank-year).
  - High-polluting industries list (Jiangsu version) includes: steel, cement, thermal power, textile, chemical, petrochemical, nonferrous metal melting, sintering pellet, ferroalloy, steel rolling, coking, coating and plating, pharmaceutical, plastic, furniture, building materials, automotive repair and maintenance.

### Key empirical findings (baseline DID and extensions)
- Default risk:
  - The default risk of high-polluting firms rose by around 0.75–0.77 percentage points after the policy implementation (Table 3 coefficients: 0.0076**, 0.0077**, 0.0075**; standard errors 0.0033, 0.0030, 0.0031).
  - Mean default rate in the full sample = 1 percent; a 0.75–0.77 percentage-point increase is equivalent to almost 80 percent increase in the default rate relative to the sample mean.
  - Difference-in-difference estimation indicates default rates of lending to high-polluting firms targeted by the Plan increased by 80 percent following policy implementation.
- Loan spreads:
  - The loan spread for high-polluting firms increases by around 1.3 percentage points after the Action Plan (Table 3 columns (4)–(6): 0.0137**, 0.0127**, 0.0135**), which is equivalent to 5.5 percent of the mean lending spread.
  - Summary implication: Banks priced some environmental regulation risk via higher spreads, but the increase in loan spreads appears insufficient relative to the considerable rise in default rates.
- Dynamic effect:
  - Interaction coefficients mostly insignificant prior to policy enforcement; the impact materializes quickly: the coefficient on the interaction term significantly rose to 0.02 on the first month after the policy’s implementation.
- Robustness and sensitivity:
  - Lagged implementation timing (30-day and 60-day lags) produces consistent Action*Treat coefficients for default (~0.0073–0.0075) and loan spread (~0.0123–0.0131).
  - Multiple-period DID (three periods: pre, interim, post) yields Action1*Treat ~0.0044–0.0046 and Action*Treat ~0.0077–0.0078 for default; loan spread Action1*Treat ~0.0113–0.0116 and Action*Treat ~0.0130–0.0138.
  - Firm-level panel (firms borrowing both before and after): Action*Treat default coefficients: 0.0027** to 0.0035***; loan spread Action*Treat: 0.0058–0.0063.
  - Same bank before and after: Action*Treat default coefficients: 0.0014* to 0.0024***; loan spread Action*Treat: 0.0033–0.0035.
  - Placebo tests not significant; Propensity Score Matching (PSM) yields Action*Treat default ~0.0058–0.0060*** and loan spread ~0.0093–0.0120***.
- Descriptive statistics (Table 2, sample 1 September 2012–31 December 2014; repayment traced to 31 March 2016):
  - Lending spread: low-polluting mean = 0.231, Std. Dev. = 0.274; high-polluting mean = 0.243, Std. Dev. = 0.239.
  - Lending rate: both groups mean = 0.074, Std. Dev. ~0.015–0.016.
  - Benchmark interest rate (percent): low-polluting mean = 5.980, Std. Dev. = 0.233; high-polluting mean = 5.955, Std. Dev. = 0.218.
  - Loan default: low-polluting mean = 0.010, Std. Dev. = 0.098; high-polluting mean = 0.010, Std. Dev. = 0.102.
  - Loan amount (CNY 10 thousand): low-polluting mean = 885, Std. Dev. = 2678; high-polluting mean = 468, Std. Dev. = 1115.
  - Maturity: short-term loans — low-polluting mean = 0.922; high-polluting mean = 0.983.
  - Loan security: secured loan share ~0.446 low-polluting and 0.445 high-polluting.
  - Firm size composition: micro and small enterprises — low-polluting 0.832; high-polluting 0.885. Big enterprises: low-polluting 0.037; high-polluting 0.020.
  - Bank types: Big five share — low-polluting 0.362; high-polluting 0.296. Rural commercial banks share — low-polluting 0.361; high-polluting 0.486.
  - Local economic structure: share of secondary industry ~0.508 (low) vs 0.511 (high); GDP per capita (CNY Yuan) mean = 96,834 (low) vs 106,863 (high).

### Heterogeneity and cross-sectional results
- Firm size:
  - Action*Treat*Medium-sized Enterprises on default: 0.0086*** (0.0025); on loan spread: 0.0431** (0.0201).
  - Action*Treat*Micro and Small Enterprises on default: 0.0158*** (0.0024); on loan spread: 0.0555* (0.0265).
  - Conclusion: largest treatment effects for small and micro enterprises.
- Firm ownership:
  - Action*Treat*SOEs on default: -0.0045** (0.0018); on loan spread: 0.0253 (0.0179) and 0.0441** (0.0204) in an alternative specification.
  - Conclusion: SOEs have relatively lower probability of default following the Plan, plausibly due to implicit or explicit government guarantees.
- Bank characteristics:
  - Action*Treat*Big Five on default: 0.0028* (0.0016); on loan spread: 0.0009 (0.0052) and 0.0087* (0.0044) in another specification.
  - Action*Treat*JECBs on default: 0.0101*** (0.0017); on loan spread: 0.0477*** (0.0047).
  - Action*Treat*Green on default: 0.0049*** (0.0015); on loan spread: 0.0136*** (0.0032).
  - Interpretation: JECBs and banks self-identified as “Green” increase lending spreads for high-polluting industries; state-owned banks underprice some transition risk relative to joint equity banks.
- Regional stringency:
  - Higher emission abatement targets for Jing-Jin-Ji, YRD, and PRD; among six prefectures, three in YRD (highly-regulated) and three outside (lightly-regulated). Results suggest stronger effects in highly-regulated cities (details reported in Appendix tables).

### Dynamics, robustness, and sensitivity
- Month-by-month interaction tracing over a 23-month window (11 months before to 12 months after implementation) finds:
  - No measurable pre-treatment differences in default rates.
  - Coefficient on interaction term rose to 0.02 in the first month after implementation.
- Robustness checks include lagged Action (30 and 60 days), multiple-period DID, alternative sample windows, firm-fixed panels (firms borrowing both before and after; same-bank samples), placebo tests, and PSM — all broadly consistent with baseline findings.
- Key control-variable effects (selected, values preserved):
  - Ln(Loan amount) on loan spread: -0.0112*** (0.0019).
  - Short-term loan on loan spread: 0.0230* (0.0125).
  - Fiduciary loan on loan spread: -0.0883*** (0.0052).
  - Loan on guarantee on loan spread: -0.0821*** (0.0090).
  - Pledged loan on loan spread: -0.1707*** (0.0175).
  - Discount loan on loan spread: -0.0888*** (0.0034).
  - Micro and small enterprises on loan spread: 0.0408** (0.0187).
  - Ln(times of borrowing) on loan spread: -0.0232*** (0.0030).
  - Number of borrowing banks on loan spread: -0.0063*** (0.0008).
  - Benchmark interest rate on default: 0.0051*** (0.0010) / 0.0055*** (0.0010) / 0.0052*** (0.0009) across specifications.

### Contributions, interpretation, and policy implications
- Contributions:
  - Links financial markets to environmental transition by incorporating environmental risk into a DSGE model with endogenous default risk.
  - Integrates environmental factors into macroeconomy through a financial accelerator mechanism.
  - Provides quantitative evidence using a large micro-level loan dataset and a quasi-natural experiment.
- Interpretation:
  - Empirical evidence aligns with model predictions: tightening environmental regulation increases firms’ production costs, lowers collateral values, increases default probabilities, and prompts banks to raise lending spreads and reduce credit supply.
  - Environmental policy targeting non-green firms can generate spillovers to green firms via bank pricing and balance-sheet channels.
- Policy recommendations:
  - An orderly and smooth transition that provides adequate time for production technology adjustment could minimize financial risks associated with environmental regulation.
  - Financial institutions should embed environmental adjustment risks into their risk management and pricing strategies.
  - Central banks and financial regulators should integrate monitoring of environment and climate change-related financial risks into prudential supervision to ensure resilience of the financial system to potential risks.

### Appendix I. — Clean Air Action Plan and banking context (selected facts)
- Clean Air Action Plan timeline and targets:
  - On 1 January 2013, the Chinese government began publishing the air quality index (AQI) in real time in 74 cities.
  - A massive fog and haze affected about 600 million people in 2013.
  - In mid-January 2013, the AQI in Beijing soared as high as 993.
  - Population-weighted mean concentration of PM2.5 for China as a whole was 54 μg/m3 in 2013.
  - On 12 September 2013, China’s State Council released the Action Plan for Air Pollution Prevention and Control.
  - By 2017, for all the second- and third-tier cities, the annual average concentration of PM2.5 should decline by at least 10 percent compared with the 2012 level.
  - Regional PM2.5 reduction targets by 2017: 25 percent for Jing-Jin-Ji, 20 percent for YRD, and 15 percent for PRD.
  - For Beijing, the annual average concentration of PM2.5 should remain at the 60 μg/m3 level.
- Industry-focused measures and financing constraints:
  - Emission intensity in key industries should reduce by over 30 percent.
  - Banks should not be allowed to provide loans to projects that have failed to pass an EIA and energy saving examinations.
  - The Plan forbids regions and industrial sectors that failed to achieve air pollution reduction goals from building new projects that would emit the same nonattainment pollutant.
- Banking industry context:
  - Total assets of the Chinese banking system amounted to 268.2 trillion yuan (or US$38.9 trillion) by the end of 2018.
  - The "Big Five" control approximately 45 percent of the market share.
  - Financial institution counts (end-2018) per CBIRC: 4,588 financial institutions, including 134 city commercial banks, 1,427 rural commercial banks, 1,616 village banks, 812 rural credit cooperative, 115 foreign banks.

*Source: Appendix I., wpiea2021228-print-pdf*

### Appendix I. ............................................................................................................

### Appendix I.

### I. Introduction and Context
- Facing escalated physical risks of environmental degradation and climate change, a growing number of countries have either tightened their existing environmental standards or imposed entirely new environmental obligations on the business community.
- Transition to low-emission and sustainable development requires a massive reallocation of capital and adoption of clean production technology, likely generating significant and systematic impacts on financial stability and macroeconomic conditions.
- Transition risks are a new category and remain relatively underexplored compared with physical risks.
- The paper investigates how an environmental policy aiming at emission abatement affects financial stability from theoretical and empirical perspectives, focusing on channels through which the transition impacts a commercial bank’s balance sheet.

### III. Theoretical framework: Environmental DSGE (E-DSGE) model
- Construct an environmental dynamic stochastic general equilibrium (E-DSGE) model to illustrate reactions of productive firms and financial institutions to policy regulations (e.g., emission caps).
- The model predicts a financial accelerator mechanism whereby tightening environmental policies lead entrepreneurs to default endogenously on the outstanding value of their debt.
- Mechanism detail:
  - Environmental policies affect the value of collateral pledged against borrowing.
  - Reduced collateral values increase endogenous default probabilities.
  - This generates an excess premium on loans (difference between contracted loan rates and risk-free rates).
- Model extends financial accelerator literature (Kiyotaki and Moore (1997), Bernanke et al. (1999), Christiano et al. (2014)) by incorporating an environmental policy aimed at reducing productive firms’ emissions.

### Empirical strategy and data
- Empirical setting: Chinese Clean Air Action Plan as a quasi-natural experiment; the year 2013 marks the start of China’s war on air pollution.
- Policy milestone: Clean Air Action Plan enacted on 12 September 2013; it set quantitative air quality improvement goals and required local governments to tighten regulation on manufacturing, especially high-emission or high-energy-consumption industries.
- Data:
  - Unique micro-level dataset containing 1.3 million loans granted to nonfinancial firms located in Jiangsu Province.
  - Borrowers classified into high-polluting firms and low-polluting firms (low-polluting firms serve as control group).
- Identification strategy: difference-in-difference (DID) using before-and-after policy variation and cross-industry variation given emission-intensity heterogeneity across industries.

### Footnote: PM definitions (preserve technical definitions)
- PM, which stands for particulate matter, contains microscopic solid particles or liquid droplets in the air. Some particles, such as dust, dirt, soot, or smoke, are large or dark enough to be seen with the naked eye. Others are so small they can only be detected using an electron microscope.
- PM10 and PM2.5 refers to inhalable particles, with diameters of 10 micrometers and smaller and 2.5 micrometers and smaller respectively.
- Particulate matter contains microscopic solids or liquid droplets that can be inhaled and cause serious health problems. The particles less than 10 micrometers in diameter can get deep into peoples’ lungs or even bloodstream, causing serious health problems.

### Key empirical findings
- Default risk:
  - The default risk of high-polluting firms rose by around 80 percent after the policy implementation.
- Loan spreads:
  - Compared with the period before the Action Plan, the loan spread for high-polluting firms increases by only around 5.5 percent.
- Bank heterogeneity in pricing transition risk:
  - Joint equity commercial banks with lower degree of government intervention and better corporate governance appropriately priced their exposure to transition risks.
  - State-owned banks failed to factor in such risks when extending credit to borrowers targeted by the environmental regulation.

### Policy implications and interpretation
- Transition toward a low-emission economy yields considerable long-term benefits, including cleaner air, improved health, reduced occurrence of natural disasters and sustainable economic growth.
- Short-run costs can be large and induce fragility in the banking sector due to compromised profitability for entrepreneurs.
- If banks underprice transition risk, financial stability may be threatened through higher realized defaults without commensurate loan pricing adjustments.
- Appropriate risk pricing and stronger corporate governance in banks can mitigate spillovers from environmental regulation to the financial sector.

### Contributions of the paper
- Links financial markets to environmental transition by incorporating environmental risk into a DSGE model with endogenous default risk.
- Integrates environmental factors into macroeconomy through a financial accelerator mechanism.
- Provides quantitative evidence of the impact of environmental regulation on financial stability using a large micro-level loan dataset and a quasi-natural experiment.

*Source: Appendix I., wpiea2021228-print-pdf*

### Section IV reports the impulse responses to a tightened environmental policy. Section V provides

### wpiea2021228 — Sections II–VI (model, impulse responses, data, empirical strategy, baseline results)

### Contribution to the literature
- Extends literature linking financial markets to climate change and environmental risk; connects transition risks to financial stability by investigating financial accelerator channels through which adjustment toward a low-emission economy affects the macroeconomy and financial sector in the short term.
- Integrates environmental risk into a DSGE model with endogenous default determined through balance-sheet changes and current asset holdings as collateral; relates to Ivanov et al. (2021), Christiano et al. (2014), Comerford and Spiganti (2017), and others cited.
- Develops a financial-accelerator mechanism that propagates environmental-risk effects by linking entrepreneurs to banks through collateral value.
- Addresses gap: empirical evidence on the financial accelerator mechanism in environmental transition is sparse; this paper uses both an E-DSGE model and a quasi-natural experiment (China’s Clean Air Action Plan) to provide theoretical and empirical evidence.

### Model (E-DSGE) — structure and key parameterization
- Core features:
  - Two types of entrepreneurs: green (g) and non-green (ng).
  - Perfectly competitive intermediate-good producers, final-goods producers, capital producer, representative household, banking sector, and a monetary authority.
  - Entrepreneurs finance capital via net worth N_e and bank loans b_e; face idiosyncratic log-normal shocks (ω) with time-varying standard deviation σω_j,t following an AR(1).
  - Environmental policy modeled as an environmental policy shock ε_M,t (AR(1)) that constrains polluting input use and reduces emissions; tightening reduces the pollution constraint and raises production costs for non-green firms.
  - Endogenous default: entrepreneurs default when ex-post return to capital is below loan repayment (cut-off ω̄ determined by Π(ω)=0).
  - Banking sector faces capital-to-asset requirement ρ_b and cost Θ(x_t) for deviating from requirement; participation constraint ensures expected loan returns cover cost of funds and losses.

- Key equations and mechanisms (referenced in text):
  - Entrepreneur profit Π(ω) = ω R_K q_k k_e − R_z b_e (Eq. 3.2).
  - Default threshold from Π(ω)=0 (Eq. 3.3).
  - Emission constraint tightened via a fixed reduction and environmental policy shock ε_M,t (Eqs. 3.8–3.9).
  - Bank participation constraint reflecting expected loan returns net of monitoring and seizure (Eq. 3.14).
  - Monetary policy: Taylor-type rule (Eq. 3.25).
  - Market clearing conditions, Calvo price setting, capital accumulation and adjustment cost functions described in the model (Eqs. 3.17–3.24, 3.21–3.22).

- Parameterization (time unit: quarters; Table 1 values preserved):
  - β Households Discount factor: 0.93
  - β_e Entrepreneurs Discount factor: 0.92
  - σ_c Elasticity of substitution for consumption: 1.353
  - ν_L Labor disutility parameter: 1
  - η Labor supply aversion: 2
  - δ_k Capital depreciation parameter: 0.025
  - ψ_k Capital adjustment cost: 6.7
  - α Capital Share: 0.46
  - γ_g Energy Inputs Share: 0.099
  - γ_ng Pollution Inputs Share: 0.09
  - θ Intensity Target coefficient: 0.05
  - F_j(ω̄_g) Probability of default (Green): 0.009
  - F_j(ω̄_ng) Probability of default (Non-Green): 0.010
  - ξ Price Elasticity of Demand for Good n: 6
  - θ Calvo’s Price Parameter for Nominal Rigidities: 0.568
  - ρ_R Monetary Policy Inertia: 0.73
  - ρ_Y Monetary Policy Reaction to Y: 0.109
  - ρ_π Monetary Policy Reaction to π: 1.487
  - β_b Banks Discount factor: 0.93
  - ρ_b Banks Capital ratio: 0.08
  - Θ Cost of deviation from the required capital ratio: 0.25
  - μ_j Monitoring Cost: 0.21
  - σ_g Size of Green Firms: 0.3
  - ρ_M Persistency of Environ. Policy shock: 0.97
  - σ_M Standard deviation on Environ. Policy shock: 0.01
- Additional calibration notes:
  - Discount factor β = β_b = 0.93 targets average Chinese one-year official lending rate of 7.4 percent over 1991–2012; entrepreneurs face lower β_e = 0.92.
  - γ_j = 0.099 set so that averaged energy expenditures account for 12 percent of GDP.
  - To achieve a 20 percent decrease in emissions, environmental policy shock persistence ρ_M = 0.97 and σ_M = 0.01.
  - Default data: average probability of default set to 0.009 (green) and 0.010 (non-green) from CBIRC corporate credit database.

### Impulse-response scenario and simulated quantitative results (Section IV)
- Scenario:
  - Government tightens environmental regulatory standards so that the pollution constraint decreases by 20 percent (simulation: one standard deviation negative shock to M; impulse responses reported as percentage deviations from steady state over a 20-quarter period).
- Simulated model findings:
  - Productivity of non-green sectors dampened by around 2 percent following the emission cap (non-green firms required to cut polluting inputs).
  - Consequences for non-green entrepreneurs:
    - Decrease in investment and fall in price of capital.
    - Lower rental return on capital.
    - Fall in collateral value reduces demand for bank funds and raises cut-off value ω̄ (higher default threshold).
  - Change in default probabilities (distributional movement illustrated in Figures 2–3):
    - Default rates increase by around 7.5 percent for the non-green sector and 1.8 percent for the green sector (model-simulated percentage increases in default rates).
  - Excess premium (spread of contractual lending rate over risk-free rate) responses:
    - Excess premium increases by around 0.5 percent for the non-green sector and 0.1 percent for the green sector.
  - Banking sector responses and spillovers:
    - Banks cut the supply of loans due to reduced bank capital (higher defaults and lower asset prices) — a deleveraging/capital channel.
    - Banks charge higher lending rates to both non-green and, to a lesser extent, green entrepreneurs to recover monitoring costs and forgone repayments — a bank-funding channel.
    - Green sector experiences negative spillovers through higher borrowing costs and lower demand for external funds, reducing clean-energy input usage and investment; some green entrepreneurs face increased default risk (smaller magnitude).
  - Summary conclusion from simulations:
    - Environmental policy targeting non-green emissions can spill over to the green sector via banking-sector pricing and balance-sheet channels, producing large negative effects on asset prices and borrowing conditions.

### Data and empirical strategy (Section V)
- Quasi-natural experiment:
  - Clean Air Action Plan (China, national enactment 12 September 2013; Jiangsu provincial launch 4 January 2014) used to evaluate short-term financial impacts of tightened environmental regulation.
- Data source:
  - Corporate credit database from China Banking and Insurance Regulatory Commission (CBIRC) Jiangsu Office.
  - Coverage: loans granted by all banks operating in six prefectures in Jiangsu province, 2010–2016 (dataset used for analysis of loans granted between 1 September 2012 and 31 December 2014; repayment status traced to 31 March 2016).
  - Sample in main analysis: 52 commercial banks; 452,208 loans to 60,856 firms (loans in Table 2 sample: 351,888 loans to low-polluting firms; 100,320 loans to high-polluting firms).
  - Loan-level details: unique firm identifier, firm fundamentals (age, size, ownership, location), bank characteristics, loan amount, maturity, credit guarantee, issuing date, delinquency status; monthly mandatory loan updates.
- Prefecture context (Jiangsu):
  - Population 80.4 million in 2018; area 102,600 km^2.
  - 2018 GDP US$1.377 trillion (RMB9.2 trillion).
  - 2013 annual industrial SO2 emissions per unit land area: 8.48 tonnes/km^2; CO2: 19.51 tonnes/km^2.
  - Total commercial bank assets in Jiangsu: RMB 16782 billion (around US$2,494 billion), 8 percent of national commercial banking industry as of 2018.
- Outcome variables:
  - Loan default indicator (1 if defaulted, 0 otherwise).
  - Loan spread measured as percentage deviation of lending rate of each loan from the benchmark rate (to remove benchmark-rate adjustments).
- Empirical design:
  - Difference-in-differences (DID) framework: treatment = borrowers in high-polluting industries targeted by Clean Air Action Plan (Jiangsu list); control = other firms.
  - Time windows:
    - Pre-regulation period: 1 September 2012–10 September 2013.
    - Interim period: 12 September 2013–4 January 2014.
    - Post-regulation period: 6 January 2014–31 December 2014.
  - Main regression (Eq. 5.1): y_l b f t = β0 + β1 Action + β2 Action*Treat + β3 Treat + controls (loan and firm characteristics, regional macro X_c) + fixed effects + u_l b f t.
  - Key identification and controls:
    - Fixed effects: time, industry, bank, prefecture; also bank*year and prefecture*year in robustness checks to control for time-varying supply-side factors and city-specific shocks.
    - Control variables: loan amount (log), maturity, loan type, firm age, ownership, size, credit rating, number of financial institutions from which firm borrows, number of times firm borrowed from bank b, firm size (proxy for monitoring incentives).
    - Standard errors clustered at industry level (robustness: cluster by bank and bank-year).
  - High-polluting industries list (Jiangsu version) includes: steel, cement, thermal power, textile, chemical, petrochemical, nonferrous metal melting, sintering pellet, ferroalloy, steel rolling, coking, coating and plating, pharmaceutical, plastic, furniture, building materials, automotive repair and maintenance.

### Empirical results — baseline DID (Section VI.A)
- Main baseline DID findings on default (columns (1)–(3) of Table 3; standard errors clustered at industry level; controls for loan and firm characteristics and fixed effects):
  - The coefficient on Action*Treat is positive and statistically significant across specifications.
  - Magnitude: default risk of high-polluting firms rose by around 0.75‐0.77 percentage points after policy implementation.
  - Comparison: mean default rate in the full sample = 1 percent; a 0.75–0.77 percentage-point increase is equivalent to almost 80 percent increase in the default rate relative to the sample mean.
- Loan-level and firm-level correlates of default:
  - Higher likelihood of default associated with:
    - Loans with shorter maturity.
    - Loans to private firms.
    - Loans to small and micro enterprises.
  - Probability of default rises with number of borrowing banks (multi-bank borrowing complicates monitoring and increases default likelihood).
- Descriptive statistics (Table 2; sample 1 September 2012–31 December 2014; repayment traced to 31 March 2016):
  - Lending spread: low-polluting mean = 0.231, Std. Dev. = 0.274; high-polluting mean = 0.243, Std. Dev. = 0.239.
  - Lending rate: both groups mean = 0.074, Std. Dev. ~0.015–0.016.
  - Benchmark interest rate (percent): low-polluting mean = 5.980, Std. Dev. = 0.233; high-polluting mean = 5.955, Std. Dev. = 0.218.
  - Loan default: low-polluting mean = 0.010, Std. Dev. = 0.098; high-polluting mean = 0.010, Std. Dev. = 0.102.
  - Loan amount (CNY 10 thousand): low-polluting mean = 885, Std. Dev. = 2678; high-polluting mean = 468, Std. Dev. = 1115.
  - Maturity: short-term loans — low-polluting mean = 0.922; high-polluting mean = 0.983.
  - Loan security: secured loan share ~0.446 low-polluting and 0.445 high-polluting.
  - Firm size composition: micro and small enterprises — low-polluting 0.832; high-polluting 0.885. Big enterprises: low-polluting 0.037; high-polluting 0.020.
  - Bank types: Big five share — low-polluting 0.362; high-polluting 0.296. Rural commercial banks share — low-polluting 0.361; high-polluting 0.486.
  - Local economic structure: share of secondary industry ~0.508 (low) vs 0.511 (high); GDP per capita (CNY Yuan) mean = 96,834 (low) vs 106,863 (high).
- Interpretation:
  - Empirical results support model implications: tightening environmental regulation increased default risk for high-polluting firms and prompted banks to adjust pricing and monitoring; this is consistent with model channels where environmental policy raises firms’ production costs, lowers collateral values, increases default probabilities, and causes banks to raise lending spreads and reduce credit supply.
  - Model and empirical evidence together indicate substantial short-term financial stability implications from environmental-transition policies, including spillovers to green firms via banking-sector reactions.

*Source: Excerpt from wpiea2021228-print-pdf (Sections II–VI) — IMF working paper (canonical PDF provided).*

### 3. The model specifications are similar to Eq. (5.1). We control for the loan and

### 3. The model specifications are similar to Eq. (5.1)

### Baseline DID findings: effect of the Clean Air Action Plan on high-polluting firms
- Main DID interaction (Action*Treat) on loan spread: positive and significant; the Clean Air Action Plan makes the lending spread to polluting firms significantly increase by 1.3 percentage points, which is equivalent to 5.5 percent of the mean lending spread.
- Main DID interaction (Action*Treat) on default:
  - Coefficients (Table 3, columns (1)–(3)): 0.0076**, 0.0077**, 0.0075** (standard errors 0.0033, 0.0030, 0.0031).
  - Interpretation: After the policy was enforced, the default probability of high-polluting firms rose by around 0.75–0.77 percentage points, or 12.6 percent to 12.9 percent compared with the benchmark interest rate of 5.975 percent (0.75 percent/5.975 percent—0.77 percent/5.975 percent).
  - Note: The interest rate shall also increase by around 14 percent to cover the increased default risk.
- Main DID interaction (Action*Treat) on loan spread (Table 3, columns (4)–(6)): 0.0137**, 0.0127**, 0.0135** (standard errors 0.0056, 0.0052, 0.0053).
- Summary implication: Banks priced some environmental regulation risk via higher spreads, but the increase in loan spreads appears insufficient relative to the considerable rise in default rates.

### Key control-variable effects (selected, values preserved)
- Ln(Loan amount) on loan spread: -0.0112*** (0.0019).
- Short-term loan on loan spread: 0.0230* (0.0125).
- Fiduciary loan on loan spread: -0.0883*** (0.0052).
- Loan on guarantee on loan spread: -0.0821*** (0.0090).
- Pledged loan on loan spread: -0.1707*** (0.0175).
- Discount loan on loan spread: -0.0888*** (0.0034).
- Micro and small enterprises on loan spread: 0.0408** (0.0187).
- Ln(times of borrowing) on loan spread: -0.0232*** (0.0030).
- Number of borrowing banks on loan spread: -0.0063*** (0.0008).
- Benchmark interest rate on default: 0.0051*** (0.0010) and on default other specifications 0.0055*** (0.0010), 0.0052*** (0.0009).

### Interpretations from controls (textual findings preserved)
- Banks tend to charge lower interest rates for loans with larger amounts and for loans with guarantees, but higher interest on small and medium sized enterprises.
- Loan spread declines as the number of borrowing banks increases, reflecting a competition effect.
- Borrowers who are private or small in size are more likely to default and pay higher borrowing rates.
- Longer bank–firm relationship (Ln(times of borrowing)) reduces default probability and lowers interest rates.
- Firms borrowing from a larger number of banks are more likely to default, but pay lower interest rates, consistent with monitoring difficulty and competition.

### Cross-sectional heterogeneity — firm characteristics (DDD results)
- Effect heterogeneity by firm size (Panel A, Table 4):
  - Action*Treat*Medium-sized Enterprises on default: 0.0086*** (0.0025); on loan spread: 0.0431** (0.0201).
  - Action*Treat*Micro and Small Enterprises on default: 0.0158*** (0.0024); on loan spread: 0.0555* (0.0265).
  - Conclusion: DDD estimates show the largest treatment effects for small and micro enterprises; impacts of environmental regulation are stronger for smaller firms.
- Effect heterogeneity by firm ownership (Panel B, Table 4):
  - Action*Treat*SOEs on default: -0.0045** (0.0018); on loan spread: 0.0253 (0.0179) and 0.0441** (0.0204) in an alternative specification.
  - Conclusion: Compared with non-SOEs, the probability of default for SOEs is relatively lower following implementation of the Clean Air Action Plan, plausibly due to implicit or explicit government guarantees.

### Cross-sectional heterogeneity — bank characteristics (DDD results)
- Bank size and ownership (Panel A, Table 5):
  - Action*Treat*Big Five on default: 0.0028* (0.0016); on loan spread: 0.0009 (0.0052) and 0.0087* (0.0044) in another specification.
  - Action*Treat*JECBs on default: 0.0101*** (0.0017); on loan spread: 0.0477*** (0.0047).
  - Interpretation: JECBs (joint equity commercial banks) significantly increase lending spreads for highly polluting industries when default rates rose; small banks may have limited capacity to raise lending rates despite higher default risk.
- Bank greenness (Panel B, Table 5):
  - Action*Treat*Green on default: 0.0049*** (0.0015); on loan spread: 0.0136*** (0.0032).
  - Action*Green on default and loan spread show positive and significant coefficients (Action*Green: 0.0036*** (0.0006); Action*Green loan spread 0.0252*** (0.0030) / 0.0264*** (0.0032) in other specifications).
  - Interpretation: “Green” banks (those reporting green lending to the CBRC) charge higher loan prices to high-polluting firms facing higher environmental policy risk.

### Regional heterogeneity (city regulation stringency)
- The Clean Air Action Plan targets varied regionally, with higher emission abatement targets for Beijing-Tianjin-Hebei, the Pearl River Delta, and the Yangtze River Delta.
- Among the six prefectures in the database, three are in the Yangtze River Delta (denoted highly-regulated cities) and three are outside (lightly-regulated).
- The interaction of the DID term with city-level stringency is used to test whether effects are stronger in highly-regulated cities (equation (6.5)); results reported in Table A4 (not reproduced here in full in the supplied content).

*Source: wpiea2021228-print-pdf - 3. The model specifications are similar to Eq. (5.1). We control for the loan and*

### Appendix 3 reports show that the lending spread has increased by a larger degree in the

### wpiea2021228-print-pdf - Appendix 3 reports show that the lending spread has increased by a larger degree in the

### Dynamics of Environmental Regulation and Financial Stability
- Estimation approach:
  - Interacted the treatment dummy with each month’s dummy to trace month-by-month effects on default.
  - Excluded the month when the Clean Air Action Plan (Jiangsu version) was enacted.
  - Considered a 23-month window spanning from 11 months before implementation to 12 months after:
    - 11 months before: 1 September 2012 to 1 September 2013
    - 12 months after: 6 January 2014 to 31 December 2014
  - Controlled for year, bank and prefecture fixed effects.
- Key dynamic finding:
  - Interaction terms between treatment and month dummies are mostly insignificant prior to policy enforcement, indicating no measurable pre-treatment differences in default rates between control and treatment groups.
  - The impact materializes quickly: the coefficient on the interaction term significantly rose to 0.02 on the first month after the policy’s implementation.
- Contextual note:
  - Default rates for clean versus dirty industries exhibited slight differences a couple of months before the national announcement due to the “PM2.5 crisis” in early 2013 and pre-emptive local actions (e.g., closures in mega cities), but the massive nationwide transition toward low-emission production, especially in manufacturing, starts after the Clean Air Action Plan came into force.

### Robustness Checks and Sensitivity Analyses
- Lagged implementation timing:
  - Lagged policy by 30 days and 60 days to account for loan negotiation/approval timing.
  - Panel A (lag 30 days) highlights (selected coefficients):
    - Action lag(30)*Treat: 0.0073 * (0.0036); 0.0073 ** (0.0031); 0.0071 ** (0.0034)
    - Loan spread (Action lag(30)*Treat): 0.0131 ** (0.0056); 0.0123 ** (0.0053); 0.0128 ** (0.0054)
    - Observations: 422929 (default), 433137 (loan spread)
    - R2: 0.019 to 0.493 across specifications
  - Panel B (lag 60 days) highlights (selected coefficients):
    - Action lag(60)*Treat: 0.0074 * (0.0037); 0.0075 ** (0.0032); 0.0072 * (0.0035)
    - Loan spread: 0.0125 ** (0.0056); 0.0119 ** (0.0052); 0.0121 ** (0.0053)
    - Observations: 410673 (default), 420090 (loan spread)
    - R2: 0.019 to 0.493 across specifications
- Multiple-period DID (three periods: pre, interim, post):
  - Panel C highlights:
    - Action1*Treat: 0.0046 *** (0.0015); 0.0044 *** (0.0013); 0.0045 ** (0.0016)
    - Action*Treat: 0.0077 ** (0.0032); 0.0078 ** (0.0028); 0.0077 ** (0.0030)
    - Loan spread (Action1*Treat): 0.0116 *** (0.0033); 0.0115 *** (0.0032); 0.0113 *** (0.0035)
    - Loan spread (Action*Treat): 0.0138 ** (0.0055); 0.0130 ** (0.0053); 0.0135 ** (0.0053)
    - Action (post-treatment dummy) and Action1 (interim dummy) also show positive and significant coefficients in many specifications.
    - Observations: 501066 (default), 523216 (loan spread)
    - R2: 0.018 to 0.500 across specifications
  - Interpretation: Coefficients for Action1*Treat are smaller than those for Action*Treat, indicating banks in Jiangsu adjusted pricing more after local enforcement than immediately after national announcement.
- Alternative sample specifications:
  - Checked different pre-policy windows (10 months: 1 September 2012 to 30 June 2013; 9 months: 1 September 2012 to 30 April 2013) for loan issuance; findings align with baseline (results available upon request).
  - Restricted analysis to loans issued before 31 Dec 2014 and traced repayment to 30 June 2015 and 31 December 2015; DID results in Table A5 (Appendix 2) consistent with baseline.
- Addressing selection/endogeneity:
  - Constructed firm-level panel of firms borrowing both before and after the Action; also restricted to firms borrowing from the same bank before and after.
  - Table 7 Panel A (firms borrowed both before and after):
    - Action*Treat (default): 0.0027 ** (0.0010); 0.0035 *** (0.0009); 0.0028 *** (0.0009)
    - Loan spread Action*Treat: 0.0063 ** (0.0023); 0.0058 ** (0.0021); 0.0061 ** (0.0023)
    - Observations: 382639 (default), 394973 (loan spread)
  - Table 7 Panel B (same bank both before and after):
    - Action*Treat (default): 0.0014 * (0.0008); 0.0024 *** (0.0007); 0.0016 ** (0.0007)
    - Loan spread Action*Treat: 0.0033 ** (0.0015); 0.0035 ** (0.0015); 0.0035 ** (0.0014)
    - Observations: 330938 (default), 337052 (loan spread)
- Placebo and matching tests:
  - Placebo (assuming policy change on 1 January 2012 for loans 1 January 2011 to 10 September 2013):
    - Action*Treat: -0.0039 (0.0033); -0.0031 (0.0026); -0.0041 (0.0033) — not significant for default
    - Loan spread placebo coefficients: -0.0014 (0.0012); 0.0011 (0.0013); -0.0004 (0.0024) — not significant
    - Observations: 589410 (default), 602693 (loan spread)
    - Interpreted as validating the causal inference.
  - Propensity Score Matching (PSM) via Kernel nearest neighbor matching:
    - Action*Treat (default): 0.0060 *** (0.0018); 0.0058 *** (0.0018); 0.0059 *** (0.0018)
    - Action*Treat (loan spread): 0.0120 *** (0.0037); 0.0093 ** (0.0039); 0.0118 *** (0.0037)
    - Observations: 198790 (default), 200632 (loan spread)
    - Consistent with baseline: default probability and lending spread for high-polluting firms both rose following implementation.

### Empirical Results and Key Statistics
- Dataset and setting:
  - Micro-level dataset contains 1.3 million commercial loans granted by all types of commercial banks operating in six Chinese prefectures to all non-financial firms.
  - All loans in main analyses granted between 1 September 2012 and 31 December 2014; repayment traced up to 31 March 2016 in many specifications.
- Main empirical magnitudes:
  - Difference-in-difference estimation indicates default rates of lending to high-polluting firms targeted by the Plan increased by 80 percent following policy implementation.
  - Loan spreads of these loans also rose, but by a much smaller degree than defaults.
  - Dynamic estimation: coefficient on interaction term rose to 0.02 in the first month after implementation.
  - Robust positive and significant Action*Treat and Action1*Treat coefficients across DID, lagged, multiple-period, panel, and PSM specifications (select examples reported in Robustness Checks).
- Heterogeneity by bank ownership and governance:
  - DDD analysis implies joint equity commercial banks with lower government intervention and better corporate governance appropriately priced exposure to transition risks.
  - State-owned banks failed to fully factor such risks when extending credit to borrowers targeted by the regulation.

### Conclusions and Policy Implications
- Interpretation:
  - Transition toward a low-emission economy is a source of structural change that significantly affects economic sectors and financial stability.
  - Empirical evidence aligns with an environmental dynamic general equilibrium (E-DSGE) model predicting higher default and lending rates when an environmental policy shift (e.g., emission cap) is imposed.
- Policy recommendations:
  - An orderly and smooth transition that provides adequate time for production technology adjustment could minimize financial risks associated with environmental regulation.
  - Financial institutions should embed environmental adjustment risks into their risk management and pricing strategies.
  - Central banks and financial regulators should integrate monitoring of environment and climate change-related financial risks into prudential supervision to ensure resilience of the financial system to potential risks.

*Source: Excerpt from wpiea2021228-print-pdf (Appendices and main text provided in source content).*

### APPENDIX I.

### APPENDIX I.

### CLEAN AIR ACTION PLAN

- The enforcement of the Clean Air Action Plan in 2013 is identified as an exogenous policy shock for this paper.
- Key timeline and triggering events:
  - On 1 January 2013, the Chinese government began publishing the air quality index (AQI) in real time in 74 cities.
  - A massive fog and haze affected about 600 million people in 2013.
  - In mid-January 2013, the AQI in Beijing soared as high as 993.
  - The population-weighted mean concentration of PM2.5 for China as a whole was 54 μg/m3 in 2013.
  - On 12 September 2013, China’s State Council released the Action Plan for Air Pollution Prevention and Control.
- Geographic focus and quantitative targets:
  - The Plan focused on three key regions: Beijing-Tianjin-Hebei (Jing-Jin-Ji), the Yangtze River Delta (YRD), and the Pearl River Delta (PRD).
  - By 2017, for all the second- and third-tier cities, the annual average concentration of PM2.5 should decline by at least 10 percent compared with the 2012 level.
  - Regional PM2.5 reduction targets by 2017: 25 percent for Jing-Jin-Ji, 20 percent for YRD, and 15 percent for PRD.
  - For Beijing, the annual average concentration of PM2.5 should remain at the 60 μg/m3 level.
- Legal and institutional reinforcement:
  - The new Air Pollution Prevention and Control Law took effect on January 1, 2016 and reinforced the Clean Air Action Plan.
  - Measures addressed pollution sources from coal, heavily polluting industries, vehicles, marine vessels and agricultural machinery, as well as the construction and food industries.
- Implementation mechanics:
  - The Plan specified key targets, strategies, and measures, often in the form of administrative orders.
  - After the national Plan, each provincial unit signed Letters of Responsibility with the Ministry of Environmental Protection and issued its own Action setting reduction goals for PM10 or PM2.5.
  - Targets and measures were specified at regional, sub-regional, sectoral, and sometimes firm levels; responsibilities were divided among governmental departments.
  - Annual implementation plans under multi-year plans specified targets, measures, and projects requiring completion within each year.
- Industry-focused measures and constraints on financing:
  - Manufacturing sectors were a primary focus, with enforced industrial upgrading and restructuring for high emission, high energy consumption, backward productivity, or excess capacity industries.
  - Strategies included end-of-pipe measures, optimizing industrial structure, promoting cleaner production and eco-industrial parks, and adjusting energy supply and consumption structure.
  - Mandatory application and upgrading of removal technologies for SO2, NO2, particulate matters, and volatile organic compounds (VOCs) in key polluting sectors.
  - Emission intensity in key industries should reduce by over 30 percent.
  - Outdated production lines and small polluting firms should close.
  - Entry requirements for highly polluting and energy-consuming sectors (iron and steel, cement, electrolytic aluminum, coking) were to be strengthened, including formulation of ban lists for construction and expansion.
  - It should be compulsory to carry out environmental impact assessment (EIA) and energy-saving examination before construction, transformation, and expansion of industrial projects.
  - Approval of an EIA should take into account total emissions of SO2, NO2, particulate matters, and VOCs as prerequisites.
  - Banks should not be allowed to provide loans to projects that have failed to pass an EIA and energy saving examinations.
  - The Plan forbids regions and industrial sectors that failed to achieve air pollution reduction goals from building new projects that would emit the same nonattainment pollutant.
  - The Plan set targets to reduce coal consumption and promote renewable energy sources.
  - Governmental departments were expected to seek policy, funding, and technological support from corresponding ministries or departments of higher-level governments; special plans for major polluting sectors could be established.

### BANKING INDUSTRY IN CHINA

- Role and scale:
  - Banks play a very important role as most Chinese firms largely rely on bank debt financing.
  - The total assets of the Chinese banking system amounted to 268.2 trillion yuan (or US$38.9 trillion) by the end of 2018.
  - This is roughly 3 times the size of the country’s annual GDP and overtakes the eurozone’s banking assets.
- Historical evolution:
  - Before 1978, the banking system was a mono-bank system with the People’s Bank of China (PBoC) functioning both as central bank and commercial bank.
  - Economic reforms since 1978 diversified the financial system.
  - In 1983, four state-owned specialized banks were established:
    - Industrial and Commercial Bank of China (ICBC) — corporate lending.
    - Agriculture Bank of China (ABC) — rural economic development.
    - Bank of China (BOC) — foreign exchange business.
    - China Construction Bank (CCB) — construction and infrastructure development.
  - PBoC’s mandate changed to that of a traditional central bank.
- Bank tiers and market shares:
  - Bank of Communications (BoCom), established in 1987, was the first joint equity bank.
  - The "Big Five" (the four state-owned banks and BoCom) are under direct control of the central government and held by the Ministry of Finance and China Investment Corporation; they control approximately 45 percent of the market share.
  - The second tier contains 12 joint equity commercial banks (JECBs), mainly state-owned but more locally oriented.
  - The third tier includes rural credit cooperatives, city commercial banks, trust and investment companies, finance companies, foreign banks, and others.
- Post-WTO reforms and regulatory changes:
  - After entry to the WTO in 2001, further reforms occurred.
  - In 2003, the Chinese Banking Regulatory Commission (CBRC) was established to monitor commercial bank operations.
  - The four state-owned banks were allowed to go public from 2005 to 2010, and city commercial banks were encouraged to bring in foreign strategic investors, go public, reconstruct, and operate across regions.
  - In 2006, the government completely opened the RMB business to foreign banks.
  - As a result, the proportion of assets of state-owned commercial banks decreased from 58.03 percent in 2003 to 37.29 percent in 2016.
  - Assets of joint-stock commercial banks increased from 10.70 percent in 2003 to 18.72 percent in 2016.
- Financial institution counts (end-2018):
  - According to China Banking and Insurance Regulatory Commission (CBIRC), there were 4,588 financial institutions by the end of 2018, including:
    - 134 city commercial banks
    - 1,427 rural commercial banks
    - 1,616 village banks
    - 812 rural credit cooperative
    - 115 foreign banks
  - (Figures used in the paragraph come from China Banking Regulatory Commission Annual Reports published in various years.)

*Source: APPENDIX I. of the provided IMF content unit.*

### References

### References

### Environmental policy, regulation, and productivity
- Albrizio, S., E. Botta, , T. Koźluk,  and V. Zipperer. 2014. “Do Environmental Policies Matter for Productivity Growth?” OECD Economic Department Working Paper 1176. doi:10.1787/5jxrjncjrcxp-en.
- Annicchiarico, B. and F. Di Dio. 2015. “Environmental Policy and Macroeconomic Dynamics in a New Keynesian Model,” Journal of Environmental Economics and Management, 69: 1–21.
- Berman, E. and L.T. Bui.  2001a. “Environmental Regulation and Productivity: Evidence from Oil Refineries,” Review of Economics and Statistics, 83 (3): 498–510.
- Berman, E. and L.T. Bui. 2001b. “Environmental Regulation and Labor Demand: Evidence from the South Coast Air Basin,” Journal of Public Economics, 79 (2): 265–295.
- Fischer, C. and M. Springborn. 2011. “Emissions Targets and the Real Business Cycle: Intensity Targets versus Caps or Taxes,” Journal of Environmental Economics and Management, 62 (3): 352–366.
- Fan, Q., X. Zhou, and T. Zhang. 2016. “Externalities of Dynamic Environmental Taxation, Paths of Accumulative Pollution and Long-term Economic Growth,” Economic Research Journal, 8: 116–128.
- Kozluk, T. and V. Zipperer. 2015. “Environmental Policies and Productivity Growth,” OECD Journal: Economic Studies, 2014 (1): 155–185.
- Tanaka, S., W. Yin, and G. H. Jefferson. 2014. “Environmental Regulation and Industrial Performance: Evidence from China,” Working Paper, Tufts Univ., Medford, MA.
- Xiao, B., Y. Fan, and X. Guo. 2018. “Exploring the Macroeconomic Fluctuations Under Different Environmental Policies in China: a DSGE Approach,” Energy Economics, 76: 439–456.
- Fan, Q., X. Zhou, and T. Zhang. 2016. “Externalities of Dynamic Environmental Taxation, Paths of Accumulative Pollution and Long-term Economic Growth,” Economic Research Journal, 8: 116–128.
- Sheehan, P., E. Cheng, A. English and F. Sun. 2014. “China’s Response to the Air Pollution Shock,” Nature Climate Change, 4: 306–309.
- Greenstone, M., 2002. “The Impacts of Environmental Regulations on Industrial Activity: Evidence from the 1970 and 1977 Clean Air Act Amendments and the Census of Manufactures,” Journal of Political Economy, 110 (6): 1175–1219.
- Greenstone, M. and R. Hanna. 2014. “Environmental Regulations, Air and Water Pollution, and Infant Mortality in India,” American Economic Review, 104 (10): 3038–3072.
- Greenstone, M., J.A. List, and C. Syverson. 2012. “The Effects of Environmental Regulation on the Competitiveness of US Manufacturing,” NBER Working Paper 18392.
- Walker, W.R. 2013. “The Transitional Costs of Sectoral Reallocation: Evidence from the Clean Air Act and the Workforce,” Quarterly Journal of Economics, 128 (4): 1787–1835.
- Berman, E. and L.T. Bui. 2001a. “Environmental Regulation and Productivity: Evidence from Oil Refineries,” Review of Economics and Statistics, 83 (3): 498–510.
- Berman, E. and L.T. Bui. 2001b. “Environmental Regulation and Labor Demand: Evidence from the South Coast Air Basin,” Journal of Public Economics, 79 (2): 265–295.

### Climate change, carbon risk, and macroeconomic implications
- Bansal, R., M. Ochoa, and D. Kiku, 2016. “Climate Change and Growth Risks,” NBER Working Paper 23009.
- Carney, M., 2015. “Breaking the Tragedy of the Horizon–climate Change and Financial Stability,” Speech given at Lloyd’s of London 29 September.
- Dafermos, Y., M. Nikolaidi, and G. Galanis. 2018. “Climate Change, Financial Stability and Monetary Policy,” Ecological Economics, 152: 219–234.
- Daniel, K.D., R.B. Litterman, and G. Wagner. 2016. “Applying Asset Pricing Theory to Calibrate the Price of Climate Risk,” NBER Working Paper 22795.
- Comerford, D. and A. Spiganti. 2017. “The Carbon Bubble: Climate Policy in a Fire-sale Model of Deleveraging,” In: 2017 Meeting Papers 734. Society for Economic Dynamics.
- de Greiff, K., M. Delis, and  S. Ongena. 2018. “Being Stranded on the Carbon Bubble? Climate Policy Risk and the Pricing of Bank Loans,” CEPR Discussion Paper No. DP12928.
- Giglio, S., M. Maggiori, J. Stroebel, and A. Weber. 2015. “Climate Change and Long-run Discount Rates: Evidence from Real Estate,” NBER Working Paper 21767.
- Daniel, K.D., R.B. Litterman, and G. Wagner. 2016. “Applying Asset Pricing Theory to Calibrate the Price of Climate Risk,” NBER Working Paper 22795.
- Mathiesen, B.V., H. Lund, and K. Karlsson. 2011. “100 Percent Renewable Energy Systems, Climate Mitigation and Economic Growth,” Applied Energy, 88 (2): 488–501.
- Trenberth, K.E., A. Dai, G. Van Der Schrier, P.D. Jones, J. Barichivich, K.R. Briffa, and J. Sheffield. 2014. “Global Warming and Changes in Drought,” Nature Climate Change, 4(1): 17.

### Financial frictions, banking, and firm finance
- Berger, A.N. and G.F. Udell. 1995. “Relationship Lending and Lines of Credit in Small Firm Finance,” Journal of Business, 68(3): 351–381.
- Bernanke, B.S., M. Gertler, and S. Gilchrist. 1999. “The financial Accelerator in a Quantitative Business Cycle Framework,” Handbook of Macroeconomics, 1: 1341–1393.
- Brandt, L. and H. Li. 2003. “Bank Discrimination in Transition Economies: Ideology, Information, or Incentives? Journal of Comparative Economics, 31 (3): 387–413.
- Cohn, J. and Deryugina, T. 2018. “Firm-Level Financial Resources and Environmental Spills,” NBER Working Paper 24516.
- Cull, R., and L.C. Xu. 2003. “Who Gets Credit? The Behavior of Bureaucrats and State Banks in Allocating Credit to Chinese State-owned Enterprises,” Journal of Development Economics, 71 (2): 533–559.
- Cull, R., W. Li, B. Sun, and L.C. Xu. 2015. “Government Connections and Financial Constraints: Evidence from a Large Representative Sample of Chinese Firms,” Journal of Corporate Finance, 32: 271–294.
- Degryse, H. and S. Ongena. 2005. “Distance, Lending Relationships, and Competition,” Journal of Finance, 60 (1): 231–266.
- Ivanov, I., M.S. Kruttli, and S.W. Watugala. 2021. “Banking on Carbon: Corporate Lending and Cap-and-Trade Policy,”  Available at SSRN 3650447: https://ssrn.com/abstract=3650447.
- Petersen, M.A. 2004. “Information: Hard and Soft,” Northwestern University Working Paper.
- Ru, H. 2018. “Government Credit, a Double-edged Sword: Evidence from the China Development Bank,”  Journal of Finance, 73 (1): 275–316.
- Schoenherr, D. 2019. “Political Connections and Allocative Distortions,” Journal of Finance, 74 (2): 543–586.
- de Greiff, K., M. Delis, and  S. Ongena. 2018. “Being Stranded on the Carbon Bubble? Climate Policy Risk and the Pricing of Bank Loans,” CEPR Discussion Paper No. DP12928.
- Kim, T. and Q. Xu. 2017. “Financial Constraints and Corporate Environmental Policies,” Available at SSRN 302876: https://ssrn.com/abstract=3028768.
- Kiyotaki, N. and J. Moore. 1997. “Credit Cycles,” Journal of Political Economy, 105 (2): 211–248.
- Iacoviello, M. 2015. “Financial Business Cycles,” Review of Economic Dynamics, 18(1): 140–163.
- Justiniano, A., G.E. Primiceri, and A. Tambalotti. 2015. “Household Leveraging and Deleveraging,” Review of Economic Dynamics, 18 (1): 3–20.
- Forlati, C. and L. Lambertini. 2011. “Risky Mortgages in a DSGE Model,” International Journal of Central Bank, 7 (1): 285–335.
- Kiyotaki, N. and J. Moore. 1997. “Credit Cycles,” Journal of Political Economy, 105 (2): 211–248.
- Daniel, K.D., R.B. Litterman, and G. Wagner. 2016. “Applying Asset Pricing Theory to Calibrate the Price of Climate Risk,” NBER Working Paper 22795.

### China-specific empirical and theoretical studies
- Bai, C.-E., C.-T. Hsieh,  and Y. Qian.2006. “The Return to Capital in China,” Brookings Papers on Economic Activity. The Brookings Institution.
- Huang, Z. 2005. “A Study on Business Cycle and the Effects of Fiscal Policy in China: an Empirical Study on a RBC Model with Three Sections,” Economic Research Journal, 6: 27–39.
- Huang, B., Y. Yang, and Z. Yu. 2020. “Strategic Environmental Regulation and Inbound Foreign Direct Investment in People’s Republic of China,” ADBI Working Paper 1129.
- Qian, Y. 1996. “Enterprise Reform in China: Agency Problems and Political Control,”  Economics of Transition and Institutional Change, 4(2): 427–447.
- Song, Z., K. Storesletten, and F. Zilibotti. 2011. “Growing Like China,” American Economic Review, 101 (1): 196–233.
- Zhao, L., X. Zhang, S. Wang, and S. Xu. 2016. “The Effects of Oil Price Shocks on Output and Inflation in China,” Energy Economics, 53: 101–110.
- Tanaka, S., W. Yin, and G. H. Jefferson. 2014. “Environmental Regulation and Industrial Performance: Evidence from China,” Working Paper, Tufts Univ., Medford, MA.
- Cull, R., and L.C. Xu. 2003. “Who Gets Credit? The Behavior of Bureaucrats and State Banks in Allocating Credit to Chinese State-owned Enterprises,” Journal of Development Economics, 71 (2): 533–559.
- Cull, R., W. Li, B. Sun, and L.C. Xu. 2015. “Government Connections and Financial Constraints: Evidence from a Large Representative Sample of Chinese Firms,” Journal of Corporate Finance, 32: 271–294.
- Ru, H. 2018. “Government Credit, a Double-edged Sword: Evidence from the China Development Bank,”  Journal of Finance, 73 (1): 275–316.
- Sheehan, P., E. Cheng, A. English and F. Sun. 2014. “China’s Response to the Air Pollution Shock,” Nature Climate Change, 4: 306–309.

### Empirical methods and econometrics
- Heckman, J., H. Ichimura, J. Smith, and P. Todd. 1998. “Characterizing Selection Bias Using Experimental Data,” Econometrica, 66 (5): 1017–1098.
- Rosenbaum, P.R., and D.B. Rubin. 1983. “The Central Role of the Propensity Score in Observational Studies for Causal Effects,” Biometrika, 70 (1): 41–55.
- Petersen, M.A. 2004. “Information: Hard and Soft,” Northwestern University Working Paper.
- Berger, A.N. and G.F. Udell. 1995. “Relationship Lending and Lines of Credit in Small Firm Finance,” Journal of Business, 68(3): 351–381.

*References section from wpiea2021228-print-pdf*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021228-print-pdf.pdf_