Structural Monetary Policy, Macroprudential Policy, and Commercial Bank Risk-Taking

Fan Yang

doi:10.4236/me.2026.176042

Modern Economy > Vol.17 No.6, June 2026

Structural Monetary Policy, Macroprudential Policy, and Commercial Bank Risk-Taking

Fan Yang
China Cinda Asset Management Co., Ltd., Beijing, China.
DOI: 10.4236/me.2026.176042 PDF HTML XML 12 Downloads 99 Views

Abstract

While actively promoting economic structural adjustments, structural monetary policy also influences commercial banks’ risk-taking. The coordination of macroprudential instruments can mitigate the risk spillover effects of structural monetary policy on commercial banks. By manually collecting collateral-related information from commercial banks and constructing key variables through textual analysis, this study employs a difference-in-differences (DID) approach to demonstrate that structural monetary policy significantly increases commercial bank risk-taking, whereas macroprudential policy effectively curbs such risks. Furthermore, heterogeneous effects are observed across different types of commercial banks and macroprudential instruments. Additionally, accommodative aggregate monetary policy amplifies both the risk spillover effects of structural monetary policy and the mitigating role of macroprudential policy.

Keywords

Structural Monetary Policy, Macroprudential Policy, Commercial Banks, Risk-Taking

Share and Cite:

Yang, F. (2026) Structural Monetary Policy, Macroprudential Policy, and Commercial Bank Risk-Taking. Modern Economy, 17, 794-831. doi: 10.4236/me.2026.176042.

1. Introduction

Since the convening of the 19th National Congress of the Communist Party of China in 2017, the “dual-pillar” regulatory framework—comprising monetary policy and macroprudential policy—has undergone continuous refinement. Distinct from monetary policy, which primarily targets price stability (“stabilizing prices”), macroprudential policy focuses on preventing major systemic risks (“stabilizing finance”; Ma & Chen, 2013; Gu & Bian, 2022). As China entered a phase of economic transformation, the central bank intensified its support for the real economy, incorporating structural adjustments into its policy objectives. Against this backdrop, structural monetary policy emerged as an innovative instrument designed to optimize resource allocation, achieving macroeconomic stability while fostering microeconomic equilibrium and sustainability.

Within this context, commercial banks—serving as both implementers and transmission channels for central bank policies—have expanded credit provision to targeted sectors (e.g. green industries, small and micro enterprises; Deng et al., 2021). However, the implementation of structural monetary policy not only alters the asset composition of commercial banks but also elevates their latent risks. On one hand, loans directed toward policy-guided sectors may increase uncertainty about asset quality (Wang & Wu, 2019); on the other, an overall expansion in lending could exacerbate credit and liquidity risks. For instance, loans extended to small and micro enterprises—characterized by default rates exceeding industry averages—inevitably raise banks’ risk exposure (Qian & Wu, 2020). Thus, analogous to the “dual-pillar” framework, structural monetary policy necessitates coordination with macroprudential instruments to mitigate risk spillovers while preserving financial stability and achieving targeted regulatory outcomes.

Examining the impact of structural monetary policy on financial stability through the lens of commercial bank risk-taking holds significant practical relevance. Risk-taking is broadly defined as the degree to which economic entities assume risks and their corresponding behavioral patterns, a concept initially explored in corporate settings (Schoemaker, 1993). Some scholars equate risk with uncertainty, where risk-taking reflects the extent of uncertainty borne by firms. In the banking sector, risk-taking encompasses uncertainties encountered during operational activities, as well as banks’ proactive risk preferences and risk-assumption behaviors in decision-making. Existing literature extensively analyzes factors influencing commercial bank risk-taking, including macroeconomic conditions, regulatory policies, market competition, corporate governance, and bank-specific characteristics (Laeven & Levine, 2008; Borio & Zhu, 2012; Fang et al., 2012; Zhang & He, 2012; Zhang et al., 2013; Barth et al., 2013; Dell’Ariccia et al., 2017).

Under the traditional “dual-pillar” framework, monetary policy influences commercial bank risk-taking through two primary channels: on the one hand, the broad credit channel, which operates via demand-side effects (i.e. altering borrowers’ balance sheets) and supply-side effects (i.e. modifying banks’ loanable funds); On the other hand, the risk-taking channel, which adjusts banks’ risk perception and tolerance, thereby shaping lending behaviors (Borio & Zhu, 2012; Fang et al., 2012; Zhang et al., 2013; Dell’Ariccia et al., 2017; Jiang & Chen, 2012; Jin et al., 2014; Albertazzi et al., 2021; Bernanke & Gertler, 1995).

Concurrently, macroprudential instruments constrain bank risk-taking through these same channels (Barth et al., 2013; Claessens et al., 2013; Shin, 2016). When categorized into loan-based, capital-based, liquidity-based, and other types, macroprudential instruments exhibit differentiated effects (Gu & Bian, 2022; Ma & Yao, 2021; Qi & Liu, 2021). Among these, loan-based and capital-based instruments demonstrate the most pronounced efficacy: the former adjusts lending constraints and transaction costs, while the latter modulates financing costs and risk tolerance, thereby mitigating monetary policy’s risk spillovers (Borio & Zhu, 2012; Yang et al., 2017).

Nevertheless, scant literature addresses the synergistic effects between structural monetary policy and macroprudential instruments. While structural monetary policy’s “targeted regulation” theoretically reduces credit distortions, empirical studies confirm that such innovative instruments still negatively impact commercial bank risk-taking (Zhou & Tao, 2019). Consequently, research on the coordination between structural monetary policy and macroprudential instruments carries both theoretical and practical significance.

This paper investigates the influence of structural monetary policy on commercial bank risk-taking from a financial stability perspective, while evaluating whether macroprudential instruments can effectively mitigate such effects. The study proceeds as follows: First, it employs the collateral expansion policy (implemented in June 2018) as a representative structural monetary instrument, applying a difference-in-differences (DID) approach to assess its impact on bank risk-taking and macroprudential policy’s moderating role. Second, it examines the credit channel’s mechanism in this process. Third, heterogeneity analyses are conducted across bank types and macroprudential instrument categories. Finally, the study explores the moderating effects of traditional monetary policy before presenting conclusions.

The main contributions include: 1) enriching literature on risk spillovers of structural monetary policy; 2) investigating policy coordination between structural monetary and macroprudential instruments; 3) validating the credit channel mechanism using DID; 4) conducting heterogeneity analyses across bank types and macroprudential instruments; 5) examining amplification effects of traditional monetary policy.

2. Research Design

2.1. Research Hypotheses

The relationship between the “dual-pillar” policy framework and commercial banks’ risk-taking constitutes a well-established academic inquiry (Ma & Chen, 2013; Gu & Bian, 2022; Borio & Zhu, 2012; Fang et al., 2012; Qi & Liu, 2021). As an innovative policy instrument, structural monetary policy shares both commonalities and distinctions with traditional monetary policy, warranting rigorous investigation into its impact on bank risk-taking and its interplay with macroprudential measures. For this study, we focus on the collateral expansion policy introduced by the People’s Bank of China in June 2018. This policy incorporated credit assets and bonds from green enterprises, small and micro enterprises (SMEs), and rural sectors into the eligible collateral framework, incentivizing commercial banks to increase loan allocations to these targeted sectors. The following analysis develops hypotheses regarding commercial bank risk-taking within this context.

The collateral expansion policy facilitates targeted credit allocation through the following operational mechanism: By designating assets from specific industries or sectors as eligible collateral, the policy enables commercial banks to pledge these eligible assets to acquire low-cost central bank refinancing funds. This incentivizes banks to expand lending to policy-targeted sectors while increasing holdings of related bond assets. Subsequently, banks may pledge these newly acquired credit and bond assets as collateral to secure additional liquidity, thereby establishing a targeted monetary transmission mechanism (Deng et al., 2021; Wang & Wu, 2019; Liu, 2016; Huang & Guo, 2021; Deng et al., 2023; Chen et al., 2024).

The collateral expansion policy primarily elevates commercial banks’ risk-taking levels by encouraging increased allocations to risk-weighted assets (Deng et al., 2021). The inclusion of targeted assets—typically from sectors like green industries, rural development, and SMEs—often falls outside banks’ optimal risk-return decision boundaries due to information asymmetry. Lending to green and small enterprises entails information friction costs and relatively higher default risks, potentially compromising bank profitability while increasing risk exposure (Hu & Zhang, 2016). Specifically, the policy expanded eligible collateral to include: 1) SME/green/rural financial bonds rated AA or above; 2) Corporate credit bonds rated AA+/AA (prioritizing SME/green economy bonds); 3) High-quality SME loans and green loans.

This incentivized banks to increase holdings of higher-risk-weight assets, altering their asset structures. Although central bank collateral provisions offer partial risk mitigation, the resultant growth in risk-weighted assets reduces capital adequacy ratios and elevates non-performing loan ratios. Furthermore, diminished profitability under structural monetary policy raises bankruptcy risks, negatively impacting Z-scores. Consequently, we propose:

H1: The collateral expansion policy significantly raises commercial banks’ risk-taking levels.

H2: This effect is transmitted through increased targeted credit allocation by commercial banks.

Analogous to the coordination between monetary and macroprudential policies in the “dual-pillar” framework, structural monetary policy can be effectively paired with macroprudential instruments to mitigate risk spillovers while achieving targeted regulatory objectives.

Three synergistic mechanisms are theorized:

Firstly, Expectation Management. Macroprudential policy signals correct risk mispricing. While collateral expansion may distort market risk assessments of certain assets (e.g., green/SME loans), macroprudential instruments inject “risk-neutral” expectations through countercyclical guidance. For instance, regulatory risk warnings can counteract irrational optimism driven by policy incentives when collateral-induced asset valuations become inflated. Simultaneously, the Macroprudential Assessment system (MPA) dynamically links banks’ risk-taking behaviors to regulatory ratings, creating intertemporal constraints that balance short-term arbitrage and long-term stability (Shin, 2016).

Secondly, Credit Demand Suppression. Macroprudential policies curb speculative financing cycles. The inclusion of SME loans as eligible collateral might encourage excessive relaxation of lending standards to obtain pledged financing quotas. Macroprudential instruments counteract this by setting SME loan concentration thresholds, compelling banks to evaluate borrowers’ actual repayment capacities when utilizing collateral policies (Adrian & Liang, 2018).

Lastly, Credit Supply Constraints. Macroprudential liquidity requirements limit asset portfolio choices. Structural monetary policy’s collateral expansion could induce structural fragility in bank balance sheets. Liquidity coverage ratios (LCR) and similar instruments prevent banks from over-relying on policy-driven maturity transformations. For example, banks channeling collateral-backed funds into targeted lending without corresponding high-quality liquid asset buffers would face deteriorating LCR metrics and rising financing costs, thereby restraining excessive holdings of low-liquidity collateral assets (Kashyap & Stein, 2000; Jiménez et al., 2012).

These mechanisms collectively form an “expectation guidance-demand suppression-supply constraint” synergy network, constraining banks’ risk decisions without impeding structural policy objectives. Thus:

H3: Macroprudential instruments effectively curb the risk-spillover effects of structural monetary policy.

Given the frequent implementation of structural monetary policies during accommodative monetary environments, examining traditional policy’s moderating effects holds practical relevance. Within the risk-taking channel framework, accommodative monetary policy compresses banks’ net interest margins by lowering risk-free rates, prompting higher risk-asset allocations to maintain profitability (Borio & Zhu, 2012). When structural and accommodative policies coincide: on the one hand, collateral expansion encourages banks to pledge more high-risk assets to the central bank for low-cost funding; on the other hand, systemic liquidity surplus depresses risk premiums, causing irrational valuations in targeted sectors (Albertazzi et al., 2021).

Correspondingly, macroprudential instruments respond through intensified interventions: 1) Loan-based instruments (e.g., sectoral loan-to-value ratio adjustments) directly constrain risk exposure growth in policy-supported sectors (Claessens et al., 2013); 2) Capital-based instruments employ countercyclical buffers to convert policy dividends into capital reserves (Shin, 2016); 3) Liquidity instruments (e.g., Net Stable Funding Ratio) mitigate maturity mismatch risks exacerbated by accommodative conditions (Adrian & Liang, 2018).

Therefore:

H4: Accommodative traditional monetary policy amplifies both the risk spillover effects of structural monetary policy and the mitigating effects of macroprudential instruments.

2.2. Difference-in-Differences (DID) Model Construction and Variable Explanation

Next, this section constructs the econometric models required for empirical analysis. Considering that this study uses microdata at the bank level, and that structural monetary policy can be regarded as exogenous to microeconomic entities to a certain extent, there is no reverse causality problem. Based on the aforementioned theoretical analysis and identification strategy, the following Difference in Difference (DID) model is constructed:

$\begin{matrix} R i s k t k_{i, t} = β_{0} + β_{1} M P_{t} D_{i} + γ^{b a n k} Z_{i, t}^{b a n k} + γ^{m a c r o} Z_{t}^{m a c r o} + μ_{i} + λ_{t} + ε_{i, t} \end{matrix}$ (1)

where i represents each sample bank, t represents time, and $β_{0}$ is the constant term. The dependent variable $R i s k t k_{i, t}$ represents the risk-taking levels of commercial banks. Risk-taking is typically measured using expected default rate, risk-weighted asset ratio, Z-score, non-performing loan rate and capital adequacy ratio. Among them, expected default rate predicts the credit risk level faced by commercial banks in businesses such as loans or investments through models. As an ex-ante risk-taking indicator, it is a relatively fair proxy variable. However, due to backward domestic credit rating data and missing data from unlisted commercial banks, it is difficult to construct this indicator; risk-weighted asset ratio measures the asset scale that commercial banks need to prepare to address risk exposure from defaults, which can relatively objectively reflect commercial banks’ risk-taking levels and is frequently used in literature (Fang et al., 2012; Jin et al., 2014); Z-score measures commercial banks’ bankruptcy risk, but since it comprehensively considers bank operating performance, leverage ratio and stability factors, it is also a relatively reasonable proxy indicator; non-performing loan rate is the ratio of non-performing loan balance to total loan balance, which can reflect the quality of banks’ credit assets (Gu & Bian, 2022; Dell’Ariccia et al., 2017; Jiang & Chen, 2012; Shin, 2016). However, considering that commercial banks can quickly reduce non-performing loan rate by increasing the total loan balance, there are problems with this proxy indicator; capital adequacy ratio directly measures risk-taking capacity, and changes in its denominator (risk assets) can reflect commercial banks’ risk-taking levels to some extent. Therefore, this paper uses risk-weighted asset ratio and Z-score as the main indicators to measure commercial banks’ risk-taking levels, while using capital adequacy ratio and non-performing loan rate as alternative indicators in robustness tests. In terms of indicator calculation, risk-weighted asset ratio (RWAR) is obtained by summing on-balance-sheet and off-balance-sheet assets with different weights (reflecting asset risk levels) and then dividing by total assets; Z-score is calculated by dividing the sum of return on assets (ROA) and capital-to-asset ratio (ETA) by the volatility of ROA, where the denominator is calculated using the rolling standard deviation of 3-year data (including 2-year lag period); non-performing loan rate (NPLR) is calculated by dividing the sum of substandard loans, doubtful loans and loss loans by the period-end total loan balance; capital adequacy ratio is calculated by dividing total capital by risk-weighted assets minus corresponding capital deductions. It should be noted that both Z-score and capital adequacy ratio are negatively correlated with risk-taking, so this paper uses their inverse forms as alternative dependent variables.

In the model, $M P_{t}$ is the time dummy variable for the implementation of structural monetary policy (the collateral expansion policy for lending facilities in June 2018). This paper uses annual frequency data, setting $M P_{t}$ to 1 for 2018 and subsequent years, and 0 for previous years; $D_{i}$ is the commercial bank grouping dummy variable. Based on the identification strategy below, commercial banks are divided into two levels according to the median of their eligible collateral ratio (the ratio of eligible collateral held by banks to total assets), with banks in the higher ratio group as the treatment group ( $D_{i} = 1$ ) and those in the lower ratio group as the control group ( $D_{i} = 0$ ). The coefficient of the interaction term $M P_{t} D_{i}$ reflects the net effect of the collateral expansion policy on commercial banks’ risk-taking levels.

In terms of identification strategy, to distinguish between the treatment group and the control group, this section uses the proportion of eligible collateral held by commercial banks as the grouping basis. Specifically, this section divides commercial banks into two groups according to the median of the proportion of eligible collateral held by each bank to total assets, with the high quantile group as the treatment group and the low quantile group as the control group. The data on eligible collateral held by commercial banks are not publicly available or directly obtainable information, and the central bank has not disclosed details of the scale of lending facilities provided to individual commercial banks. In view of this, this section refers to relevant literature and adopts a comprehensive method of database retrieval and manual data collection to systematically sort out the required eligible collateral data (Deng et al., 2021).

Regarding the scope of eligible collateral for commercial banks, before 2018, it was mainly high-credit-grade bond assets, including government bonds, central bank bills, policy financial bonds, local government bonds and AAA-grade corporate credit bonds. Since June 2018, the central bank has lowered collateral requirements for specific industries and sectors, including financial bonds of not lower than AA grade for small and micro enterprises, green and agriculture-related sectors, corporate credit bonds of AA+ and AA grades (priority given to bonds involving small and micro enterprises and green economy), as well as high-quality loans to small and micro enterprises and green loans. Based on BankFocus financial institution database, CSMAR bank research database and other public information, this section adopts the following methods to obtain data related to commercial banks’ eligible collateral: use the balance sheet (including supplementary statements) of BankFocus financial institution database to obtain information about commercial banks’ holdings of policy financial bonds, local government bonds and corporate credit bonds. However, some bond asset data in this database have serious missing problems, which need to be supplemented by CSMAR bank research database, commercial banks’ annual reports and social responsibility reports. At the same time, information about government bonds, central bank bills and corporate bonds of AA grade and above in the above reports also needs to be sorted out.

In terms of collateral-related data processing, this section adopts the following methods: 1) To accurately match the requirements of the collateral framework, only high-credit-grade bond assets are selected as collateral before 2018, and corporate credit bonds of AA grade and above are included since 2018; 2) Due to relatively scarce stock data of green loans and small and micro loans, they are temporarily excluded from statistics; 3) Some commercial banks lack detailed rating data of corporate credit bonds, making it impossible to separate high-credit-grade bonds. This chapter estimates based on the proportion information of bond ratings of banks of similar size in the same year; 4) Some banks have problems such as mixed bond classification, inconsistent caliber or even missing data in some years. This chapter makes up for them according to the proportion information of the same bank in previous years. It should be noted that due to partial missing data problems in the above data, there is a non-negligible discrepancy between the collateral statistics and the scope defined by the central bank. Therefore, this paper uses the proportion of commercial banks’ borrowing from the central bank to total assets as another criterion for dividing the treatment group and the control group in robustness analysis, and similarly determines the sample groups with higher and lower proportions of borrowing from the central bank according to quantiles. The rationality of this identification strategy is based on the following two points: On the one hand, this paper finds that the lending facilities obtained by commercial banks through collateral pledge account for a large proportion of borrowing from the central bank by comparing the collateral and borrowing from the central bank of several commercial banks; On the other hand, relevant literature proves that the lending facility policy can promote the increase of commercial banks’ borrowing from the central bank (Deng et al., 2021). Therefore, this paper uses this identification strategy to verify the robustness of the previous conclusions.

$Z_{i, t}^{b a n k}$ represents bank-level control variables, $Z_{t}^{m a c r o}$ represents macroeconomic control variables, $μ_{i}$ is the dummy variable for individual fixed effects, and $λ_{t}$ is the annual dummy variable for time fixed effects. Among them, bank-level control variables include bank size (Size), return on assets (ROA), loan-to-deposit ratio (LDR), cost-to-income ratio (CIR) and proportion of non-interest income (PNII); macro-level control variables include GDP growth rate (GDP_yoy), money supply M2 growth rate (M2_yoy) and consumer price index CPI growth rate (CPI_yoy). On this basis, this section incorporates macroprudential policy and constructs the following model:

$\begin{matrix} R i s k t k_{i, t} = β_{0} + β_{1} M P_{t} D_{i} M P P_{t} + β_{2} M P_{t} D_{i} + γ^{b a n k} Z_{i, t}^{b a n k} + γ^{m a c r o} Z_{t}^{m a c r o} + μ_{i} + λ_{t} + ε_{i, t} \end{matrix}$ (2)

where $M P P_{t}$ represents macroprudential policy, using the macroprudential index as the proxy variable of macroprudential policy (Claessens et al., 2013). With reference to relevant literature, this section constructs this index based on the iMaPP database provided by the International Monetary Fund (IMF) and the public policy documents issued by the central bank and the National Financial Regulatory Administration (formerly the CBIRC) (Gu & Bian, 2022; Qi & Liu, 2021). Specifically, first sort out the implementation of China’s macroprudential Instruments (see Table 1), and set corresponding dummy variables for each policy. When the policy instrument has no change, the value is 0, when it takes effect or tightens, the value is 1, and when it fails or relaxes, the value is −1; second, sum the variable values of each macroprudential policy from the starting point of the sample interval to obtain the corresponding policy index; finally, sum all indices to obtain the total macroprudential index, where a higher value indicates stricter macroprudential policy.

Table 1. Classification and description of macroprudential instruments.

Category	Instrument	Description
Capital-based	Countercyclical capital buffer	A systemic risk control mechanism designed to mitigate financial risks arising from economic cyclical fluctuations and sudden shocks
	Reserve capital buffer	Capital requirements set for different asset categories according to Basel III
	Capital requirements	Banking capital regulations including risk weights, systemic risk buffers, and minimum capital requirements
	Leverage ratio	Restrictions on bank leverage usage
	Requirements for systemically important banks	Measures implemented to reduce risks posed by critical financial institutions, such as additional capital and liquidity fees
Loan-based	Loan loss provisions	Reserves that financial institutions must establish for potential loan losses, including sector-specific provisions and dynamic provisions
	Loan growth ceilings	Restrictions on the increment or total amount of overall credit, corporate credit, and household credit
	Loan threshold restrictions	Restrictions on loans based on loan characteristics (e.g., size, maturity, value-to-loan ratio, interest rate type) and borrower characteristics
	Foreign currency loan constraints	Restrictions on foreign currency loans according to foreign currency loan rules
	Loan-to-value ratio (LTV)	Restrictions on the ratio of loan amount to collateral value
	Loan-to-income ratio (LTI)	Restrictions on the ratio of loan amount to borrower income
	Taxes	Taxes levied on specific transactions, assets, or liabilities
Liquidity- based	Liquidity coverage ratio (LCR)	Requirements including liquidity asset ratios, coverage ratios, core funding ratios, net stable funding ratio floors, and foreign debt restrictions
	Loan-to-deposit ratio	Restrictions on the ratio of loans to deposits
	Foreign exchange position limits	Restrictions on foreign exchange positions, risk exposures, funds, and currency mismatch regulations
	Reserve requirements	Reserve requirements set for macroprudential purposes
Other	-	Other macroprudential instruments not included in the above categories, such as structural measures, stress tests, and profit distribution restrictions

The macroprudential index constructed via the above scoring rule followed by summation strictly complies with the official compilation standard of IMF’s iMaPP database, which has been widely adopted in mainstream international literature for measuring the tightness of macroprudential policies. In practice, macroprudential instruments cover dozens of subdivided tools with heterogeneous applicable scenarios, transmission channels and regulatory effectiveness. No globally recognized differentiated weighting framework has been established to quantify heterogeneous marginal effects of various regulatory tools so far. Therefore, the equal-weighted summation method is reasonable under current theoretical and practical limitations.

Further, drawing on existing literature, macroprudential policies are classified into capital, loan, liquidity and other types according to their types, so as to construct macroprudential sub-dimension indices, which are recorded as MPP_C, MPP_Ln, MPP_Lq and MPP_E respectively (Gu & Bian, 2022; Ma & Yao, 2021). Table 1 shows the classification and description table of macroprudential instruments.

It should be noted that since all models involved in this paper include individual fixed effects and time fixed effects, the coefficients of $M P_{t} D_{i}$ and $M P P_{t}$ items will be absorbed, so their regression coefficient estimates are no longer displayed. At the same time, for policies (such as other monetary policies, fiscal policies and industrial policies) and events that may interfere with the test results during the sample period, if they can be added to the model as control items in the form of dummy variables, their impact on the dependent variable will also be absorbed by the above two fixed effects, so they are no longer displayed in the model. See Table 2 for specific variable definitions and calculation methods.

2.3. Data Sources and Descriptive Statistics

The data on commercial banks used in this paper are sourced from the BankFocus banking database and the CSMAR database, with selected annual data from balance sheets, income statements, credit risk reports, market risk reports, liquidity

Table 2. Variable definitions and calculations.

Category	Variable	Symbol	Definition
Dependent variables	Risk-weighted asset ratio (%)	$R W A R$	Risk-weighted assets/Total assets. Higher values indicate greater risk-taking
	Z-score (inverse)	$Z_r$	$σ_{R O A} / (R O A + E T A)$ , where ETA is capital-to-asset ratio. Higher values indicate greater risk-taking
	Capital adequacy ratio (inverse)	$C A R_r$	Risk-weighted assets/(Total capital − corresponding capital deductions). Higher values indicate greater risk-taking
	Non-performing loan ratio (%)	$N P L R$	(Substandard loans + Doubtful loans + Loss loans)/Total loan balance. Higher values indicate greater risk-taking
Grouping variables	Eligible collateral ratio	$C o l l t r$	Eligible collateral/Total assets
	Central bank borrowing ratio	$L o a n_C B$	Borrowings from central bank/Total assets
Explanatory variables	Collateral expansion policy	$M P_{t}$	Dummy variable: 1 after June 2018 policy implementation, 0 otherwise
	Commercial bank grouping variable	$D_{i}$	Dummy variable: 1 for high collateral ratio banks (treatment group), 0 for low ratio banks (control group)
	Macroprudential policy	$M P P$	Macroprudential policy composite index
		$M P P_C$	Capital-based macroprudential policy index
		$M P P_L n$	Loan-based macroprudential policy index
		$M P P_L q$	Liquidity-based macroprudential policy index
		$M P P_E$	Other macroprudential policy index
Channel variable	Targeted loan ratio (%)	$P L o a n$	Targeted loans issued by banks/Loans to listed enterprises
Moderating variables	Bank type	$T y p e$	State-owned, joint-stock, city, rural, and foreign banks
	Traditional monetary policy	$T M P$	Quantity-based instrument: M2 growth rate; Price-based instrument: 7-day interbank offered rate (inverse)
Bank-level controls	Bank size	$S i z e$	Logarithm of total assets
	Return on assets (%)	$R O A$	Net profit/Total assets
	Loan-to-deposit ratio (%)	$L D R$	Total loans/Total deposits
	Cost-to-income ratio (%)	$C I R$	Operating expenses/Interest income
	Non-interest income ratio (%)	$P N I I$	(Net fee & commission income + Other operating income + Investment income excluding bond interest)/Operating income
Macro-level controls	GDP growth rate (%)	$G D P_y o y$	Year-on-year GDP growth rate
	M2 growth rate (%)	$M 2_y o y$	Year-on-year M2 money supply growth rate
	CPI growth rate (%)	$C P I_y o y$	Year-on-year consumer price index growth rate

risk reports, and loan analysis reports used to construct the dependent variables and firm-level control variables. Data related to structural monetary policy and macroeconomics are also sourced from the CSMAR database. The statistics on the scale of eligible collateral held by commercial banks required for this study are obtained from the supplementary balance sheets in BankFocus, with missing data supplemented from commercial bank annual reports, social responsibility reports, and other public information. Through data collection and processing procedures, this paper compiles data from 85 commercial banks, including 6 large commercial banks, 12 joint-stock commercial banks, 49 city commercial banks, 11 rural commercial banks, and 7 foreign banks. Considering that the lending facility policy has been implemented since 2013 and some statistics have lagging characteristics, this paper selects the period from 2013 to 2022 as the sample interval, using 2018 as the DID policy time point, with 5 years before the policy and 4 years after the policy.

The final sample covering 85 commercial banks and more than 800 bank-year observations is relatively large among existing similar empirical studies on structural monetary policy and bank risk-taking. This research requires manually sorting detailed eligible collateral data, extracting non-standard information from annual and social responsibility reports via textual analysis, and matching granular loan-by-loan data of listed firms, which brings higher data collection barriers than conventional banking empirical research. Detailed micro-indicators of numerous small-and-medium-sized banks are undisclosed publicly, restricting further sample expansion.

The loan data of listed enterprises are also sourced from the CSMAR database, selecting daily loan-by-loan data from 2013 to 2022, which are then compiled and merged into loan data from major banks to listed enterprises, and matching enterprise announcements and industry information from the listed enterprise database.

The macroprudential data are sourced from the iMaPP database of the International Monetary Fund (IMF), which contains monthly data on macroprudential policies in major countries. This section aggregates the monthly frequency to annual frequency based on the starting point of the sample interval.

The processing of the original data in this paper is as follows: 1) Exclude samples with severe missing bank or enterprise financial data; 2) Exclude credit sources from non-bank financial institutions such as enterprise finance companies; 3) Exclude credit data from financial industries, delisted companies, and ST companies; 4) Winsorize all continuous variables at the 1st and 99th percentiles.

Table 3 presents the descriptive statistics of the main variables in this paper. It can be seen that among the four dependent variables involved in this paper, the risk-weighted assets (RWAR) have the largest fluctuations; the proportion of targeted loan issuance also shows a large range and variance, which may be related to the term structure of loans, as some targeted loans have longer terms and lower frequency.

Table 3. Descriptive statistics of main variables.

VARIABLES	Sample Size	Mean	Standard Deviation	Minimum	Maximum
$R W A R$	832	39.110	17.482	0.311	82.135
$Z_r$	832	75.023	6.015	13.910	153.685
$C A R_r$	832	5.178	1.012	3.057	9.468
$N P L R$	831	1.675	0.845	0.220	4.860
$C o l l t r$	832	21.754	25.852	0	74.631
$M P P$	838	14.236	5.492	0	25.000
$M P P_C$	838	4.713	3.186	0	8.000
$M P P$ _Ln	838	2.291	1.217	0	5.000
$M P P_L q$	838	4.378	3.194	0	9.000
$M P P_E$	838	0.611	0.125	0	1.000
$L o a n_C B$	813	2.302	3.606	0	64.800
$P L o a n$	784	12.317	13.349	0	69.956
$S i z e$	831	24.870	0.968	21.675	29.848
$R O A$	828	0.845	0.496	0.250	1.529
$L D R$	823	63.949	12.382	26.273	110.786
$C I R$	813	38.520	7.269	21.090	64.520
$P N I I$	796	27.300	11.778	0.865	75.630
$M 2_y o y$	838	10.496	1.957	8.100	13.600
$G D P_y o y$	838	6.026	1.900	2.200	8.100
$C P I_y o y$	838	1.687	0.692	0.891	2.904

3. Empirical Analysis

3.1. Benchmark Regression Analysis

First, to verify the changes in commercial banks’ risk-taking before and after the implementation of the collateral expansion policy, this section uses the risk-weighted asset ratio (RWAR) and the inverse Z-score (Z_r) as dependent variables and employs an event-study design to conduct the parallel-trend test illustrated in Figure 1. Among them, the period before the policy implementation is used as the base period and thus omitted, so four periods are retained on the left side of the policy implementation point in the figure. It can be seen that relative to banks holding a lower proportion of eligible collateral, the risk-taking level of commercial banks holding a higher proportion of eligible collateral shows significant changes before and after the policy implementation, passing the parallel trend test.

This section continues to conduct empirical analysis of benchmark regression using the DID model, and the results are shown in Table 4. Among them, columns (1) and (2) use the risk-weighted asset ratio (RWAR) as the dependent variable, and columns (3) and (4) use the inverse Z-score (Z_r). The core explanatory variable is $M P_{t} D_{i}$ , which represents the net effect of the collateral expansion policy. The odd-numbered columns show the regression results without control variables, while the even-numbered columns include bank-level and macro-level control variables. All models include individual fixed effects and time fixed effects.

It can be seen that after the implementation of the collateral expansion policy, commercial banks holding a higher proportion of eligible collateral show a significant increase in both the risk-weighted asset ratio and the inverse Z-score. This indicates that this policy significantly increases the risk-taking level of commercial banks, which is consistent with Hypothesis 1 of this paper. In addition, in the multivariate regression, the coefficients of bank size (Size) and return on assets (ROA) are both less than 0 and partially significant, indicating that larger size and higher profitability reduce risk-taking.

It should be noted that model (1) controls for the $D_{i}$ term, but the effect of this term is absorbed by the individual fixed effects, so it is not reflected in Table 4. Similarly, the time fixed effects also absorb the $M P_{t}$ term, as well as the effects of other monetary policies, fiscal policies, and industrial policies during the same period, and are no longer displayed in the results.

Figure 1. Parallel trend test chart for benchmark analysis.

Table 4. Benchmark regression results.

	(1)	(2)	(3)	(4)	(5)	(6)
VARIABLES	$R W A R$	$R W A R$	$R W A R$	$Z_r$	$Z_r$	$Z_r$
$M P_{t} D_{i}$	2.734***	0.996**	1.013**	0.592**	0.316**	0.327**
	(3.7697)	(2.5336)	(2.5102)	(2.1831)	(1.9884)	(2.0175)
$M P_{t} D_{i} M P P_{t}$			−0.368***			−0.175**
			(−2.6713)			(−2.3964)
$S i z e$		−1.790	−1.787		−0.424*	−0.424*
		(−1.3295)	(−1.2482)		(−1.7914)	(−1.7359)
$R O A$		−6.728	−6.726		−1.822***	−1.813***
		(−1.4803)	(−1.4851)		(−2.6931)	(−2.6694)
$L D R$		1.128*	1.128*		0.012	0.012
		(1.7825)	(1.7978)		(0.9629)	(0.9315)
$C I R$		0.473	0.474		0.005	0.006
		(0.6032)	(0.6495)		(0.0928)	(0.0887)
PNII		0.019	0.020		0.006	0.008
		(1.5201)	(1.5148)		(1.1830)	(1.1845)
$M 2_y o y$		−0.072	−0.069		−0.208	−0.204
		(−0.3167)	(−0.3086)		(−1.5074)	(−1.4842)
$G D P_y o y$		0.698	0.699		−0.060	−0.059
		(1.5299)	(1.4973)		(−1.3107)	(−1.3126)
$C P I_y o y$		−1.657	−1.653		−0.095	−0.095
		(−0.6893)	(−0.6234)		(−1.3065)	(−1.3137)
$C o n s t a n t$	38.257***	19.484	22.149	0.119***	1.562**	1.634**
	(5.6024)	(1.4105)	(1.4687)	(3.4956)	(1.9807)	(1.9681)
$I n d i v i d u a l F E$	√	√	√	√	√	√
$T i m e F E$	√	√	√	√	√	√
$O b s e r v a t i o n s$	832	787	787	832	787	787
$R - s q u a r e d$	0.076	0.172	0.174	0.168	0.313	0.315
$N u m b e r o f i d$	82	79	79	82	79	79

Note: * indicates significant at the 10% level, ** indicates significant at the 5% level, and *** indicates significant at the 1% level. The numbers in parentheses are t values.

3.2. Channel Test

Next, this section further examines the transmission channels through which the collateral expansion policy affects commercial banks’ risk-taking. According to the previous analysis, the collateral expansion policy increases commercial banks’ risk-taking levels by inducing them to increase risk-weighted assets. Therefore, this section constructs the following model based on Equation (1):

$\begin{matrix} R i s k t k_{i, t} = β_{0} + β_{1} M P_{t} D_{i} C h a n l_{i, t} + β_{2} M P_{t} D_{i} + β_{3} D_{i} C h a n l_{i, t} + β_{4} M P_{t} C h a n l_{i, t} \\ + β_{5} C h a n l_{i, t} + γ^{b a n k} Z_{i, t}^{b a n k} + γ^{m a c r o} Z_{t}^{m a c r o} + μ_{i} + λ_{t} + ε_{i, t} \end{matrix}$ (3)

where, $C h a n l_{i, t}$ is the channel variable of interest in this section, measured by the proportion of “targeted” loans issued by commercial banks to total loans, denoted as PLoan. Here, “targeted” refers to specific industries or sectors covered by the collateral framework. Considering that the primary targeted regulatory functions of the lending facility policy focus on green, small and micro, and agriculture-related enterprises, this section compiles annual loan volumes for each bank from the loan-by-loan credit data of listed enterprises, and then identifies loans to these three types of enterprises. Specifically, for green enterprise loans, this section defines green projects based on the “Green Credit Statistical Form Instructions” issued by the China Banking Regulatory Commission in 2013, constructs a green keyword library using text analysis methods, and thoroughly compares the keywords with all companies’ annual announcements in 2018. Loans to enterprises that implemented energy-saving and environmental protection projects during the year are identified as green enterprise loans. For small and micro enterprise loans, this section refers to the “Statistical Classification Standards for Large, Medium, Small and Micro Enterprises (2017)” (National Statistics [2017] No. 213), setting different ranges of operating income and number of employees for different industries. Enterprises meeting the small and micro standards are assigned a value of 1 for $D_{i}$ , otherwise 0 (for example, the classification standard for small and micro enterprises in the industrial sector is operating income below RMB 20 million and number of employees below 300). Considering the serious lack of data on employees, this section mainly classifies based on operating income and identifies loans to qualifying enterprises as small and micro enterprise loans. For agriculture-related enterprise loans, this section refers to the “Guidelines for Industry Classification of Listed Companies (2012 Revision)”, identifying loans to enterprises in industries A01 and A05 as agriculture-related enterprise loans. It should be noted that A01 represents agriculture, and A05 represents agriculture, forestry, animal husbandry, and fishery services. The latter includes services such as agricultural irrigation, primary processing of agricultural products, agricultural technology promotion, agricultural scientific research, agricultural machinery leasing, and agricultural information consulting, which are important components of the agricultural industry chain and provide necessary support and services for agricultural production and rural economic development, and are therefore also counted as agriculture-related enterprises.

Combining the previous theoretical analysis, it can be seen that the collateral expansion policy increases commercial banks’ risk-taking levels by increasing targeted loan issuance. Therefore, as the proportion of targeted loans increases, the impact of the collateral expansion policy on commercial banks’ risk-taking levels should correspondingly increase. In other words, the share of targeted lending positively moderates the net effect of the collateral expansion policy.

Columns (1) and (2) of Table 5 show the results of the mechanism test using the risk-weighted asset ratio as the dependent variable, where column (1) shows the regression results without the channel variable, and column (2) includes the proportion of targeted loan issuance. Similarly, columns (3) to (4) show the moderating variables using the inverse Z-score as the dependent variable, with the same arrangement, which will not be repeated here.

Table 5. Channel test of structural monetary policy on commercial banks’ risk-taking.

	$R W A R$		$Z_r$
	(1)	(2)	(3)	(4)
VARIABLES	$/$	$P L o a n$	$/$	$P L o a n$
$M P_{t} D_{i} C h a n l_{i, t}$		0.432**		0.118**
		(2.0258)		(1.9813)
$C h a n l_{i, t}$		1.087**		0.206**
		(1.9972)		(1.9864)
$M P_{t} D_{i}$	0.996**	0.797**	0.316**	0.247*
	(2.5336)	(2.1019)	(1.9884)	(1.7983)
$D_{i} C h a n l_{i, t}$		−0.162*		−0.031
		(−1.6742)		(−1.5962)
$M P_{t} C h a n l_{i, t}$		−0.233		−0.009
		(−0.7933)		(−0.1057)
$C o n s t a n t$	19.484	14.691	1.562**	1.014
	(1.4105)	(0.9064)	(1.9807)	(1.6316)
$B a n k C o n t r o l s$	√	√	√	√
$M a c r o C o n t r o l s$	√	√	√	√
$I n d i v i d u a l F E$	√	√	√	√
$T i m e F E$	√	√	√	√
$O b s e r v a t i o n s$	787	778	787	777
$R - s q u a r e d$	0.172	0.173	0.313	0.313
$N u m b e r o f i d$	79	78	79	78

Note: * indicates significant at the 10% level, ** indicates significant at the 5% level, and *** indicates significant at the 1% level. The numbers in parentheses are t values.

From the Table 5, it can be seen that after adding the channel variable, the collateral expansion policy still has a significant effect on commercial banks holding a higher proportion of collateral. On this basis, the proportion of targeted loan issuance by commercial banks significantly increases their risk-taking levels. At the same time, the coefficient of the triple interaction term is significantly positive, indicating that the targeted credit assets held by commercial banks strengthen the impact of the collateral expansion policy on their risk-taking. In other words, with the implementation of the collateral expansion policy, commercial banks’ targeted credit assets increase, and thus the net effect of this policy on commercial banks’ risk-taking also increases.

Furthermore, this section constructs the following model to examine the role of channel variables when structural monetary policy and macroprudential policy are implemented together:

$\begin{matrix} R i s k t k_{i, t} = β_{0} + β_{1} M P_{t} D_{i} M P P_{t} C h a n l_{i, t} + β_{2} M P_{t} D_{i} M P P_{t} + β_{3} D_{i} C h a n l_{i, t} + β_{4} M P_{t} C h a n l_{i, t} \\ + β_{5} M P P_{t} C h a n l_{i, t} + β_{6} C h a n l_{i, t} + γ^{b a n k} Z_{i, t}^{b a n k} + γ^{m a c r o} Z_{t}^{m a c r o} + μ_{i} + λ_{t} + ε_{i, t} \end{matrix}$ (4)

According to the previous analysis, structural monetary policy can increase commercial banks’ risk-taking levels by increasing their targeted credit issuance. On this basis, macroprudential policy can weaken the impact of structural monetary policy on commercial banks’ risk-taking while optimizing this credit channel mechanism. Specifically, as loan-based macroprudential instruments impose restrictions on loan thresholds, quality, and ceilings, commercial banks’ credit issuance to relatively high-risk areas such as green, small and micro, and agriculture-related sectors is strictly regulated. Commercial banks, on the one hand, need to implement the central bank’s monetary policy intentions, and on the other hand, use strict risk identification and management methods to selectively support relatively excellent and promising enterprises in these areas. Therefore, macroprudential policy is of great significance for reducing the negative effects of structural monetary policy and optimizing its transmission mechanism.

Table 6 shows the corresponding empirical results. Among them, columns (1) and (2) show the results of the mechanism test using the risk-weighted asset ratio as the dependent variable, and columns (3) and (4) use the inverse Z-score as the dependent variable. The odd-numbered columns do not include the channel variable, while the even-numbered columns include the proportion of targeted loan issuance. It can be found that, on the one hand, for commercial banks, credit to targeted areas, as relatively high-risk assets, can increase their risk-taking levels; on the other hand, the collateral expansion policy increases commercial banks’ risk-taking levels by increasing their targeted credit issuance, while macroprudential policy effectively curbs this negative effect while optimizing its credit channel.

Table 6. Channel test of structural monetary policy and macroprudential policy.

	$R W A R$		$Z_r$
	(1)	(2)	(3)	(4)
VARIABLES	$/$	$P L o a n$	$/$	$P L o a n$
$M P_{t} D_{i} M P P_{t} C h a n l_{i, t}$		−0.106**		−0.087**
		(−2.0219)		(−1.9945)
$M P_{t} D_{i} M P P_{t}$	−0.368***	−0.342***	−0.175**	−0.154**
	(−2.6713)	(−2.5943)	(−2.3964)	(−2.2869)
$D_{i} C h a n l_{i, t}$		0.059		0.026
		(0.3916)		(0.1127)
$M P_{t} C h a n l_{i, t}$		0.037		0.015
		(0.0854)		(0.0586)
$M P P_{t} C h a n l_{i, t}$		−0.007		0.012
		(−0.0031)		(0.1044)
$C h a n l_{i, t}$		0.231*		0.097**
		(1.8934)		(1.9816)
$C o n s t a n t$	22.149	20.672	1.634**	1.835**
	(1.4687)	(1.5293)	(1.9681)	(1.9695)
$B a n k C o n t r o l s$	√	√	√	√
$M a c r o C o n t r o l s$	√	√	√	√
$I n d i v i d u a l F E$	√	√	√	√
$T i m e F E$	√	√	√	√
$O b s e r v a t i o n s$	787	773	787	773
$R - s q u a r e d$	0.174	0.175	0.315	0.315
$N u m b e r o f i d$	79	78	79	78

Note: * indicates significant at the 10% level, ** indicates significant at the 5% level, and *** indicates significant at the 1% level. The numbers in parentheses are t values.

3.3. Robustness Tests

(I) Alternative Dependent Variables

According to previous literature, there are multiple ways to measure commercial banks’ risk-taking levels. The previous section presented empirical analysis results using the risk-weighted asset ratio (RWAR) and the inverse Z-score (Z_r) as dependent variables. Below, this section will use capital adequacy ratio (CAR) and non-performing loan ratio (NPLR) to proxy risk-taking. Higher capital adequacy ratios indicate lower risk-taking levels, while non-performing loan ratios change in the same direction as commercial banks’ risk-taking levels. For ease of analysis, the inverse of capital adequacy ratio (CAR_r) is used here. The parallel trend test after replacing the dependent variables is shown in Figure 2, and the benchmark regression results are shown in Table 7.

From Figure 2, it can be seen that before the implementation of the collateral expansion policy, there was no significant difference in capital adequacy ratios and non-performing loan ratios between commercial banks holding higher proportions of eligible collateral and the control group samples. However, significant changes occurred after the policy implementation, satisfying the parallel-trend assumption.

On this basis, this section presents the benchmark regression and mechanism test results after replacing the dependent variables, as detailed in Table 7. From columns (1) and (3), it can be found that structural monetary policy significantly

Figure 2. Parallel trend test after replacing the dependent variables.

Table 7. Robustness test results after replacing the dependent variables 1.

	$C A R_r$		$N P L R$
	(1)	(2)	(3)	(4)
VARIABLES	$/$	$P L o a n$	$/$	$P L o a n$
$M P_{t} D_{i}$	0.305**	0.180**	0.069**	0.043*
	(2.1674)	(1.9649)	(1.9864)	(1.8715)
$M P_{t} D_{i} C h a n l_{i, t}$		0.245**		0.085*
		(2.1627)		(1.7106)
$C h a n l_{i, t}$		0.589**		0.104*
		(1.9878)		(1.6594)
$D_{i} C h a n l_{i, t}$		−0.032		−0.014
		(−1.3886)		(−1.4436)
$M P_{t} C h a n l_{i, t}$		−0.011		−0.005
		(−0.5348)		(−0.2637)
$C o n s t a n t$	1.917	1.387	0.742	0.429
	(1.5862)	(1.6125)	(1.1945)	(0.9358)
$B a n k C o n t r o l s$	√	√	√	√
$M a c r o C o n t r o l s$	√	√	√	√
$I n d i v i d u a l F E$	√	√	√	√
$T i m e F E$	√	√	√	√
$O b s e r v a t i o n s$	786	777	781	775
$R - s q u a r e d$	0.126	0.127	0.109	0.109
$N u m b e r o f i d$	79	78	79	78

Note: * indicates significant at the 10% level, ** indicates significant at the 5% level, and *** indicates significant at the 1% level. The numbers in parentheses are t values.

reduces commercial banks’ capital adequacy ratios while increasing their non-performing loan ratios. The results in columns (2) and (4) show that as commercial banks increase their targeted loan issuance, the impact of the collateral expansion policy on banks’ risk-taking levels also increases, which is consistent with the previous conclusions.

Meanwhile, after incorporating macroprudential policy, the empirical results are shown in Table 8. It can be seen that after replacing the dependent variables, the effect of structural monetary policy on commercial banks’ risk-taking remains significant, the channel effect of targeted loan issuance remains unchanged, and macroprudential instruments continue to effectively curb risks.

Table 8. Robustness test results of alternative dependent variables 2.

	$C A R_r$		$N P L R$
	(1)	(2)	(3)	(4)
VARIABLES	$/$	$P L o a n$	$/$	$P L o a n$
$M P_{t} D_{i} M P P_{t}$	−0.215**	−0.209**	−0.107**	−0.104**
	(−2.1849)	(−2.1567)	(−1.9832)	(−1.9946)
$M P_{t} D_{i} M P P_{t} C h a n l_{i, t}$		−0.064**		−0.045**
		(−2.2195)		(−2.3418)
$C h a n l_{i, t}$		0.194**		0.076*
		(2.0137)		(1.8159)
$D_{i} C h a n l_{i, t}$		0.013		0.007
		(0.0694)		(0.0053)
$M P_{t} C h a n l_{i, t}$		0.042		0.025
		(0.1985)		(0.1397)
$M P P_{t} C h a n l_{i, t}$		−0.006		0.011
		(−0.0612)		(0.2054)
$C o n s t a n t$	1.294	1.037	0.863	0.782
	(1.4638)	(1.5175)	(1.2534)	(1.3776)
$B a n k C o n t r o l s$	√	√	√	√
$M a c r o C o n t r o l s$	√	√	√	√
$I n d i v i d u a l F E$	√	√	√	√
$T i m e F E$	√	√	√	√
$O b s e r v a t i o n s$	786	777	781	775
$R - s q u a r e d$	0.128	0.129	0.111	0.111
$N u m b e r o f i d$	79	78	79	78

Note: * indicates significant at the 10% level, ** indicates significant at the 5% level, and *** indicates significant at the 1% level. The numbers in parentheses are t values.

(II) Alternative Macroprudential Indicators

This section refers to relevant literature and constructs a “strong regulation” macroprudential dummy variable, which is set to 1 when the macroprudential composite index is above the mean and 0 otherwise. Based on this, a new macroprudential index is constructed (Gu & Bian, 2022; Jin & Jiang, 2020). The regression analysis of Model 2 is then reperformed, with results shown in columns (1) and (2) of Table 9. It can be seen that after replacing the macroprudential indicators, the inhibitory effect of macroprudential instruments on the effects of structural monetary policy remains significant.

Table 9. Robustness test results of alternative explanatory variables and channel variables.

	Replacing the Explanatory Variables		Replacing the Channel Variables
	(1)	(2)	(3)	(4)
VARIABLES	$R W A R$	$Z_r$	$R W A R$	$Z_r$
$M P_{t} D_{i} M P P_{t}$	−0.285**	−0.138**	−0.315**	−0.105**
	(−2.5429)	(−2.1734)	(−2.5037)	(−2.3496)
$M P_{t} D_{i} M P P_{t} C h a n l_{i, t}$			−0.134**	−0.102**
			(−1.9864)	(−1.9932)
$C h a n l_{i, t}$			0.254**	0.086*
			(1.9618)	(1.8943)
$D_{i} C h a n l_{i, t}$			0.067	0.031
			(0.0492)	(0.1473)
$M P_{t} C h a n l_{i, t}$			0.029	0.009
			(0.0384)	(0.0016)
$M P P_{t} C h a n l_{i, t}$			−0.011	0.005
			(−0.0024)	(0.0113)
$C o n s t a n t$	21.534	1.406*	20.247	1.933**
	(1.2955)	(1.9517)	(1.3563)	(1.9687)
$B a n k C o n t r o l s$	√	√	√	√
$M a c r o C o n t r o l s$	√	√	√	√
$I n d i v i d u a l F E$	√	√	√	√
$T i m e F E$	√	√	√	√
$O b s e r v a t i o n s$	787	787	675	676
$R - s q u a r e d$	0.174	0.315	0.173	0.313
$N u m b e r o f i d$	79	79	68	68

Note: * indicates significant at the 10% level, ** indicates significant at the 5% level, and *** indicates significant at the 1% level. The numbers in parentheses are t values.

(III) Alternative Mechanism Indicators

Next, this section attempts to reconstruct the targeted loan issuance ratio variable based on commercial banks’ public information. Specifically:

For green enterprise loans, green loan balance data (from the CSMAR database) are used. This data only covers major national state-owned and joint-stock commercial banks, resulting in a smaller sample size.

For small and micro loans, data on small and micro enterprise credit issuance by commercial banks are manually collected from public texts such as annual reports and social responsibility reports, obtaining major commercial banks’ small and micro enterprise loan balance data. It should be noted that most commercial banks ceased disclosing SME-loan data since 2017 or 2018, instead beginning to disclose inclusive small and micro enterprise credit balances, which differ from traditional small and micro enterprise credit balances in statistical caliber. Small and micro enterprise credit is mainly classified according to the “Statistical Classification Standards for Large, Medium, Small and Micro Enterprises” (National Statistics [2011] No. 75), while inclusive small and micro enterprise credit defines loans with single-household credit lines below 10 million yuan as inclusive small and micro loans (the central bank raised this recognition standard to 20 million yuan in January 2024, but this change has no impact on research within the sample period). Considering that the former has a higher overlap with the small and micro credit of concern in this paper, it is used as small and micro enterprise loans. Since most commercial banks disclosed both small and micro enterprise loan balances under different statistical calibers in the first year they began disclosing inclusive small and micro loan data, this section fills in small and micro loan balance data after 2018 based on the ratio derived from different statistical definitions for the same commercial bank during the same period. For special cases, the following treatments are made: 1) For missing data in individual years for a few samples, linear interpolation is used to fill the gaps; 2) For bank samples where the ratio of small and micro loan balances under different statistical calibers during the same period cannot be obtained, the average ratio of commercial banks of similar size is used as the standard for data completion. In addition, agriculture-related enterprise loans are not considered here due to their small scale and serious data deficiencies. This results in new targeted loan data and the reconstruction of targeted loan ratio indicators.

Based on these alternative indicators, this section reperforms the mechanism regression analysis, with results shown in columns (3) and (4) of Table 9. It can be seen that under the alternative indicators of channel variables, the collateral expansion policy can still increase commercial banks’ risk-taking levels by increasing the proportion of targeted loan issuance. At the same time, macroprudential instruments can optimize this channel, effectively mitigating the negative effects of structural monetary policy.

(IV) Alternative Identification Strategies

Considering the limitations of the current identification strategy and grouping scheme, this section replaces the identification strategy in two ways: First, the proportion of commercial banks’ borrowing from the central bank to total assets is used as the grouping standard for regrouping, with the high-level group as the treatment group and the low-level group as the control group, thus obtaining samples that are more and less affected by the policy, respectively. Second, considering the limitations of the median method, this section attempts to adopt tercile grouping, setting the samples with the highest collateral proportions as the treatment group and those with the lowest collateral proportions as the control group, while discarding the middle group samples. Under these two new identification strategies, this section performs DID model regression for benchmark regression and mechanism analysis, obtaining the test results shown in Table 10.

The first four columns of Table 10 show the regression results using the proportion of commercial banks’ borrowing from the central bank as the grouping standard, where the first two columns are based on Model 1 for regression analysis, and the latter two columns are based on Model 2 for regression analysis. The odd-numbered columns use the risk-weighted asset ratio as the dependent variable, and the even-numbered columns use the inverse Z-score as the dependent variable. Columns (5) to (8) show the test results using terciles as the grouping standard, arranged in the same way as the first four columns, which will not be repeated here. It can be seen that under these two identification strategies, the impact of structural monetary policy on bank risk-taking remains significant, and macroprudential policy can also effectively curb its negative effects. Overall, although the several identification strategies used in this paper all have certain limitations, the empirical conclusions remain unchanged, demonstrating certain robustness.

Table 10. Robustness test results under the new identification strategies.

	Strategy 1				Strategy 2
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
VARIABLES	$R W A R$	$Z_r$	$R W A R$	$Z_r$	$R W A R$	$Z_r$	$R W A R$	$Z_r$
$M P_{t} D_{i}$	0.568**	0.242*	0.612**	0.239*	0.245**	0.063*	0.249**	0.068*
	(1.9624)	(1.7918)	(1.9756)	(1.7658)	(1.9617)	(1.6953)	(1.9798)	(1.6987)
$M P_{t} D_{i} M P P_{t}$			−0.285**	−0.147**			−0.128**	−0.104**
			(−2.3604)	(−2.1315)			(−1.9701)	(−1.9842)
$C o n s t a n t$	21.537*	1.433	21.967	1.619	18.265	0.716	18.763	0.796
	(1.7821)	(0.8593)	(1.6325)	(0.9641)	(1.3328)	(0.3194)	(1.4107)	(0.3429)
$B a n k C o n t r o l s$	√	√	√	√	√	√	√	√
$M a c r o C o n t r o l s$	√	√	√	√	√	√	√	√
$I n d i v i d u a l F E$	√	√	√	√	√	√	√	√
$T i m e F E$	√	√	√	√	√	√	√	√
$O b s e r v a t i o n s$	787	787	787	787	524	524	524	524
$R - s q u a r e d$	0.165	0.271	0.167	0.273	0.149	0.243	0.151	0.245
$N u m b e r o f i d$	79	79	79	79	52	52	52	52

Note: * indicates significant at the 10% level, ** indicates significant at the 5% level, and *** indicates significant at the 1% level. The numbers in parentheses are t values.

(V) Placebo Test

Considering that the aforementioned DID test may be interfered with by other unobserved factors, leading to biased results, this section uses the placebo method to test the robustness of the previous conclusions. Specifically, this section adopts two methods for testing: The first is to set false policy shocks before the real shocks to observe whether the net effect of the collateral expansion policy on commercial banks’ risk-taking remains significant. If not, it indicates that false shocks cannot affect the conclusions of this paper, and real shocks are the true cause of the policy effects. The second is to randomly select samples as treatment and control groups, observing the coefficient estimates of the DID interaction term after multiple repeated selections. If they significantly deviate from the benchmark regression results and approach 0, it indicates that the identification strategy selection in this paper is correct, and the research results are robust and unaffected by other unobserved factors.

In the first method, this section sets false policy shock time points in 2016 and 2017, respectively, and re-estimates using the DID method, with results shown in Table 11. The first four columns in the table show the test results with false shocks set in 2016, where the first two columns are regressions based on Model 1, and the latter two columns are based on Model 2. Columns (5) to (8) show the test results under false policy shocks in 2017. It can be found that under false policy shocks, the policy no longer has a significant impact on commercial banks’ risk-taking, and the effect of macroprudential instruments is no longer significant. This indicates, to some extent, that the real policy shock in 2018 has significant implications for changes in commercial banks’ risk-taking.

In the second method, this section randomly selects treatment and control groups, observing the distribution of DID coefficient estimates after 500 repeated selections, with results shown in Figure 3. The two graphs use the risk-weighted asset ratio (RWAR) and the inverse Z-score (Z_r) as dependent variables, respectively, with the horizontal axis representing the values of DID coefficient estimates, the vertical axis representing relative probability density, the dashed line representing 0, and the solid line representing the estimated value of the DID coefficient in the benchmark regression. It can be seen that the DID estimates under random selection are all distributed around 0 and far from the benchmark regression estimates. This indicates that the impact of the collateral expansion policy on commercial banks’ risk-taking is not affected by unobserved factors.

Table 11. Robustness test results of the false policy shocks.

	False Policy Shock in 2016				False Policy Shock in 2017
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
VARIABLES	$R W A R$	$Z_r$	$R W A R$	$Z_r$	$R W A R$	$Z_r$	$R W A R$	$Z_r$
$M P_{t} D_{i}$	0.042	0.003	0.064	0.003	0.094	0.004	0.098	0.005
	(0.5216)	(0.2754)	(0.5809)	(0.3187)	(0.9197)	(1.3389)	(1.0316)	(1.3870)
$M P_{t} D_{i} M P P_{t}$			−0.038	−0.029			−0.063	−0.051
			(−1.0195)	(−1.0097)			(−1.2068)	(−1.4132)
$C o n s t a n t$	19.941	1.317	18.549	1.186	21.869	1.570	20.197	1.309
	(0.3275)	(0.9715)	(0.4196)	(0.9457)	(1.0137)	(0.8731)	(1.1964)	(0.9326)
$B a n k C o n t r o l s$	√	√	√	√	√	√	√	√
$M a c r o C o n t r o l s$	√	√	√	√	√	√	√	√
$I n d i v i d u a l F E$	√	√	√	√	√	√	√	√
$T i m e F E$	√	√	√	√	√	√	√	√
$O b s e r v a t i o n s$	787	787	787	787	787	787	787	787
$R - s q u a r e d$	0.151	0.284	0.151	0.284	0.152	0.285	0.152	0.285
$N u m b e r o f i d$	79	79	79	79	79	79	79	79

Note: * indicates significant at the 10% level, ** indicates significant at the 5% level, and *** indicates significant at the 1% level. The numbers in parentheses are t values.

Figure 3. Placebo test results of the randomly selected groups of treatment and control.

It should be noted that this method only applies to the DID benchmark analysis represented by Model 1. Since Model 2 is built on the basis of Model 1, as the latter passes the placebo test, neither will be affected by unobserved factors.

(VI) Excluding the Impact of State Capital Injections

The large-scale state capital injection events (hereinafter referred to as “state capital injections”) since February 2019 may also affect the empirical results of this paper. In February 2019, to alleviate financing difficulties and high financing costs for private enterprises, state capital began injecting funds into private enterprises. This event may affect commercial banks’ risk-taking levels through the following channels: On the one hand, state capital injections reduce the default probability of private enterprises and increase the proportion of safe assets for commercial banks, thereby reducing their risk asset ratios and non-performing loan ratios. On the other hand, state capital injection policies focus on improving the financing environment for small and medium-sized private enterprises, which will help promote commercial banks’ credit issuance to small and micro enterprises. While enhancing commercial banks’ profitability, this may also increase their risk asset scale, thereby increasing their risk-taking levels.

At the same time, state capital injections may have heterogeneous effects on different types of commercial banks. Large state-owned and joint-stock commercial banks have larger scales and more flexible adaptability, enabling them to more effectively utilize the policy dividends brought by state capital injections. City and rural commercial banks have relatively smaller capital scales and face certain limitations in capital allocation and credit issuance. In addition, foreign commercial banks may generally remain cautious and take a wait-and-see attitude toward the private enterprise credit market.

Based on this, this section uses a time dummy variable to control for the impact of state capital injections, setting its value to 0 before 2019 and 1 thereafter. Further, considering the heterogeneous effects of state capital injections on different types of commercial banks, the interaction term between this variable and bank type dummy variables is added to the model as a control variable. The empirical results are shown in Table 12.

From Table 12, it can be seen that after controlling for the impact of state capital injection events, the impact of structural monetary policy on commercial banks’ risk-taking levels decreases slightly but remains significant, while the effect of macroprudential instruments is not significantly affected.

Table 12. Robustness test results after controlling the impact of state capital injections.

	$/$		Macroprudential
	(1)	(2)	(3)	(4)
VARIABLES	$R W A R$	$Z_r$	$R W A R$	$Z_r$
$M P_{t} D_{i}$	1.015**	0.342**	1.021**	0.353**
	(2.3846)	(1.9925)	(2.4907)	(1.9945)
$M P_{t} D_{i} M P P_{t}$			−0.326***	−0.148**
			(−2.6980)	(−2.2586)
$C o n s t a n t$	17.956	1.389*	18.033	1.574*
	(1.3969)	(1.8543)	(1.4591)	(1.6932)
$B a n k C o n t r o l s$	√	√	√	√
$M a c r o C o n t r o l s$	√	√	√	√
$I n d i v i d u a l F E$	√	√	√	√
$T i m e F E$	√	√	√	√
$O b s e r v a t i o n s$	787	787	787	787
$R - s q u a r e d$	0.174	0.314	0.175	0.316
$N u m b e r o f i d$	79	79	79	79

Note: * indicates significant at the 10% level, ** indicates significant at the 5% level, and *** indicates significant at the 1% level. The numbers in parentheses are t values.

(VII) Excluding the Impact of the COVID-19 Pandemic

In January 2020, the World Health Organization officially named COVID-19 (Coronavirus Disease 2019) as the disease caused by the novel virus that triggered a surge of cases in a short period. From a theoretical perspective, the spread of the COVID-19 pandemic (hereinafter referred to as the “pandemic”) may affect commercial banks’ risk-taking levels as follows. On the one hand, the debt servicing capacity of enterprises and individuals is adversely impacted, leading to higher default risks. The asset quality of commercial banks deteriorates accordingly, elevating their risk-taking levels. On the other hand, commercial banks have to shift to conservative investment strategies in response to the low-interest-rate environment and market uncertainties. They increase holdings of government bonds and high-grade credit bonds, and expand cash and liquid asset positions to meet potential funding needs and cope with market fluctuations. As a result, their risk-taking levels may decline.

At the same time, the pandemic may have heterogeneous effects on different types of commercial banks. Large state-owned and joint-stock commercial banks, with their capital scale and policy advantages, actively adjust their asset structures to support the recovery of the real economy. City and rural commercial banks, with the support of local government policies, increase support for local and rural economies but face higher risk pressures. Foreign banks adopt more cautious strategies to maintain stable operations.

Based on this, this section attempts to control for the possible impact of the pandemic, using a time dummy variable set to 0 before 2020 and 1 thereafter. At the same time, considering that different types of commercial banks may be differentially affected, the interaction term between the pandemic time dummy variable and bank type dummy variables is also added to the model as a control variable.

The empirical results are shown in Table 13. It can be found that after controlling for the impact of the pandemic, the effect of structural monetary policy on commercial banks’ risk-taking levels remains significant, and the inhibitory effect of macroprudential policy is unaffected. Overall, the research conclusions of this paper are robust.

Table 13. Robustness test results after controlling the impact of COVID-19 pandemic.

	/		Macroprudential
	(1)	(2)	(3)	(4)
VARIABLES	$R W A R$	$Z_r$	$R W A R$	$Z_r$
$M P_{t} D_{i}$	0.913**	0.298**	0.974**	0.308**
	(2.5624)	(2.0069)	(2.4986)	(1.9983)
$M P_{t} D_{i} M P P_{t}$			−0.364***	−0.182**
			(−2.6810)	(−2.3959)
$C o n s t a n t$	17.534	1.208*	18.012	1.548*
	(1.4407)	(1.9142)	(1.4329)	(1.8753)
$B a n k C o n t r o l s$	√	√	√	√
$M a c r o C o n t r o l s$	√	√	√	√
$I n d i v i d u a l F E$	√	√	√	√
$T i m e F E$	√	√	√	√
$O b s e r v a t i o n s$	787	787	787	787
$R - s q u a r e d$	0.174	0.314	0.175	0.316
$N u m b e r o f i d$	79	79	79	79

Note: * indicates significant at the 10% level, ** indicates significant at the 5% level, and *** indicates significant at the 1% level. The numbers in parentheses are t values.

4. Further Analysis

4.1. Heterogeneity Analysis by Bank Type

For different types of commercial banks, the impacts of structural monetary policy and macroprudential instruments may exhibit heterogeneous characteristics. On one hand, from the perspective of business structure, city banks rely more heavily on credit to policy-targeted sectors. Constrained by their regionalized operations, these banks naturally tilt their credit resources toward specific policy-supported industries such as local SMEs, agriculture-related sectors, and high-tech enterprises (Borio & Zabai, 2018). Due to their homogeneous customer base, these banks are more likely to significantly increase credit allocation to targeted sectors under policy incentives, leading to further concentration of credit supply in these areas and creating a positive-feedback loop of policy “incentives → credit concentration → risk accumulation” (Jiménez et al., 2012).

In this context, macroprudential instruments can constrain this phenomenon by setting industry loan concentration thresholds and customer risk exposure limits. When the growth rate of targeted credit driven by policy exceeds the regulatory tolerance range, the concentration score in the Macroprudential Assessment system (MPA) will trigger regulatory intervention, forcing banks to reduce risk exposure through various means (Shin, 2016). Clearly, this constraint has a more significant impact on city commercial banks with relatively homogeneous business structures compared to large state-owned and joint-stock banks with diversified customers and businesses. It is worth mentioning that although rural commercial banks share similar characteristics, their smaller capital scale and service scope, mainly targeting local agricultural enterprises, make them relatively less affected by structural monetary policy and macroprudential instruments.

On the other hand, from the perspective of liquidity funds, compared with large state-owned commercial banks, joint-stock and city commercial banks have higher financing costs and fewer financing channels, resulting in weaker stability on the liability side and thus greater reliance on the central bank’s lending facilities (Kashyap & Stein, 2000). Compared with rural commercial banks, these two types of banks have broader service coverage and more channels to obtain collateral assets, enabling them to fully utilize the financing channels provided by policy instruments to alleviate maturity mismatch pressures, thus showing higher sensitivity to structural monetary policy.

In this context, macroprudential liquidity regulation instruments can effectively constrain the risk-taking behavior of these two types of banks. Specifically, compared with large state-owned commercial banks, joint-stock and city commercial banks have stronger market orientation and profit motives, but at the same time, to meet liquidity coverage ratio requirements, they need to allocate part of the funds obtained through policy instruments to maintain relatively low-yield high-liquidity asset scales, thereby constraining their risk asset expansion behavior (Adrian & Liang, 2018).

This section examines the heterogeneous effects of structural monetary policy on different types of commercial banks through empirical analysis, constructing the following model. In Equation (5), $T y p e_{i}$ is a series of 0 - 1 dummy variables representing the types of commercial banks, including large state-owned, joint-stock, city, rural, and foreign banks.

$\begin{matrix} R i s k t k_{i, t} = β_{0} + β_{1} M P_{t} D_{i} T y p e_{i} + β_{2} M P_{t} D_{i} + β_{3} M P_{t} T y p e_{i} \\ + γ^{b a n k} Z_{i, t}^{b a n k} + γ^{m a c r o} Z_{t}^{m a c r o} + μ_{i} + λ_{t} + ε_{i, t} \end{matrix}$ (5)

Further, based on Model 2 and Equation 5, this section introduces macroprudential policy to explore the impact of bank types on its policy effects:

$\begin{matrix} R i s k t k_{i, t} = β_{0} + β_{1} M P_{t} D_{i} M P P_{t} T y p e_{i} + β_{2} M P_{t} D_{i} M P P_{t} + β_{3} M P_{t} D_{i} \\ + γ^{b a n k} Z_{i, t}^{b a n k} + γ^{m a c r o} Z_{t}^{m a c r o} + μ_{i} + λ_{t} + ε_{i, t} \end{matrix}$ (6)

The test results for five types of commercial banks—large state-owned, joint-stock, city, rural, and foreign—are shown in Table 14. From columns (1) and (2), it can be seen that the coefficients of the triple interaction term MPtDi·Typei are both positive and negative. Specifically, compared with other types of banks, joint-stock and city commercial banks are more significantly affected by policy changes; from columns (3) and (4), it can be found that the effects of macroprudential instruments are also more pronounced in joint-stock and city commercial banks.

Overall, large state-owned commercial banks have rich business types, diversified asset allocation, and a more complete risk supervision system. While enjoying the benefits of structural monetary policy, they will not experience drastic changes in risk-taking levels due to targeted changes in asset allocation. Joint-stock commercial banks have stronger regulatory avoidance and profit motives. The low-cost funds from structural monetary policy may exacerbate their risk preferences, so macroprudential instruments have stronger constraints on them. City commercial banks are more regionalized and localized. After increasing credit support for local SMEs and agriculture-related enterprises, their risk-taking levels are higher, and macroprudential policies thus play a “stabilizer” role. In addition, rural commercial banks mainly serve local agricultural enterprises, and foreign banks are relatively cautious, both being less sensitive to policy impacts and less affected.

Table 14. Results of heterogeneity analysis by bank type.

		$/$		Macroprudential
		(1)	(2)	(3)	(4)
	VARIABLES	$R W A R$	$Z_r$	$R W A R$	$Z_r$
Large state-owned	$M P_{t} D_{i} T y p e_{i}$	0.098	0.045
		(1.0582)	(1.4067)
	$M P_{t} D_{i} M P P_{t} T y p e_{i}$			−0.015	−0.009
				(−0.2976)	(−0.0528)
Joint-stock	$M P_{t} D_{i} T y p e_{i}$	0.086*	0.042
		(1.6829)	(1.6398)
	$M P_{t} D_{i} M P P_{t} T y p e_{i}$			−0.117*	−0.097*
				(−1.9302)	(−1.7864)
City	$M P_{t} D_{i} T y p e_{i}$	0.104**	0.083*
		(1.9784)	(1.7395)
	$M P_{t} D_{i} M P P_{t} T y p e_{i}$			−0.142*	−0.056*
				(−1.8836)	(−1.6929)
Rural	$M P_{t} D_{i} T y p e_{i}$	−0.075	−0.104
		(−0.5842)	(−1.1965)
	$M P_{t} D_{i} M P P_{t} T y p e_{i}$			0.021	0.018
				(0.0831)	(0.0547)
Foreign	$M P_{t} D_{i} T y p e_{i}$	−0.012	−0.004
		(−0.1067)	(−0.0073)
	$M P_{t} D_{i} M P P_{t} T y p e_{i}$			0.033	0.014
				(0.0134)	(0.0089)

Note: * indicates significant at the 10% level, ** indicates significant at the 5% level, and *** indicates significant at the 1% level. The numbers in parentheses are t values.

4.2. Heterogeneity Analysis across Macroprudential Instruments

Among macroprudential instruments, different types of instruments may have heterogeneous effects on mitigating the risk spillovers of structural monetary policy. First, loan-based instruments may be the most effective. Structural monetary policy changes the balance sheet structure of commercial banks through targeted liquidity supply mechanisms, inducing them to increase risk exposure in specific sectors (e.g., SME and green industries) (Borio & Zhu, 2012). Loan-based instruments (such as industry loan concentration limits and loan-to-value ratio adjustments for specific sectors) directly enhance the sensitivity of risk pricing by adjusting the risk weights of relevant loans, thereby constraining the direction and intensity of credit allocation and achieving precise blocking of risk transmission chains (Claessens et al., 2013). The targeted intervention characteristics of these instruments give them forward-looking features in risk identification and immediate responsiveness in policy implementation.

Second, capital-based instruments exhibit weaker efficacy. Compared with the precise adjustment of loan-based instruments, capital-based instruments focus more on lagging mitigation of systemic risks (Claessens et al., 2013; Laeven & Levine, 2008). When structural monetary policy increases the risk-taking level of commercial banks, capital-based instruments adjust risk preferences by requiring countercyclical capital buffers and leverage ratio constraints, prompting banks to adjust risk preferences under capital cost constraints. However, capital-based instruments may perform worse than loan-based instruments in two aspects: On the one hand, their global characteristics may prompt commercial banks to contract credit supply in non-policy-supported sectors to meet capital requirements, thereby exacerbating risk concentration in targeted sectors (Dell’Ariccia et al., 2017); on the other hand, delays in capital replenishment decisions (i.e., commercial banks delaying regulatory impacts through profit retention or external financing) and lagging signal transmission in risk pricing (i.e., capital constraints need to adjust asset allocation behavior through market expectation adjustments) may weaken the precision and immediacy of these instruments in mitigating the risk spillover effects of structural monetary policy (Adrian & Liang, 2018).

Finally, liquidity-based instruments may have relatively weaker inhibitory effects. Liquidity coverage ratio (LCR) and net stable funding ratio (NSFR) mainly constrain the maturity transformation behavior of commercial banks, primarily targeting maturity mismatch risks rather than credit risk transmission mechanisms (Shin, 2016). The implementation of structural monetary policy enhances the ability of commercial banks to convert liquidity into credit assets in specific sectors, while liquidity-based instruments cannot identify the actual risks of these assets, making it difficult to achieve effective targeted constraints.

This section conducts empirical analysis based on Model 2, using the macroprudential sub-dimension indices constructed earlier, with results shown in Table 15.

Table 15. Results of heterogeneity analysis across macroprudential instruments.

		(1)	(2)	(3)	(4)
	VARIABLES	$R W A R$	$Z_r$	$C A R_r$	$N P L R$
Composite	$M P_{t} D_{i} M P P_{t}$	−0.368***	−0.175**	−0.215**	−0.107**
		(−2.6713)	(−2.3964)	(−2.1849)	(−1.9832)
Capital-based	$M P_{t} D_{i} M P P_C_{t}$	−0.153*	−0.136*	−0.198**	−0.095
		(−1.7426)	(−1.8621)	(−1.9835)	(−1.6478)
Loan-based	$M P_{t} D_{i} M P P_L n_{t}$	−0.379***	−0.194**	−0.221**	−0.128**
		(−2.7291)	(−2.3158)	(−2.0376)	(−1.9852)
Liquidity-based	$M P_{t} D_{i} M P P_L q_{t}$	−0.227*	−0.097*	−0.074	−0.051
		(−1.6853)	(−1.7290)	(−1.6328)	(−1.6412)
Other	$M P_{t} D_{i} M P P_E_{t}$	−0.058	−0.031	−0.062	−0.029
		(−0.9471)	(−0.4964)	(−0.2107)	(−0.0842)

Note: * indicates significant at the 10% level, ** indicates significant at the 5% level, and *** indicates significant at the 1% level. The numbers in parentheses are t values.

From Table 15, it can be found that among various macroprudential policies, loan-based instruments have the most significant effect on structural monetary policy, followed by capital-based instruments, then liquidity-based instruments, and finally other types. Specifically, loan-based instruments impose greater constraints on commercial banks’ specific asset allocation through restrictions on loan loss provisions, growth ceilings, and thresholds; capital-based instruments further inhibit the growth of commercial banks’ risk-taking through requirements for capital and leverage ratios. In addition, liquidity-based instruments can prevent excessive funds from flowing to relatively high-risk sectors through regulations on liquidity coverage ratios and loan-to-deposit ratios. In summary, macroprudential policy can effectively mitigate the impact of structural monetary policy on commercial banks’ risk-taking, and the coordinated implementation of both helps the central bank achieve its targeted regulatory objectives while maintaining economic and financial stability.

4.3. Moderating Effects of Traditional Monetary Policy

Considering that structural monetary policy has been implemented multiple times in an accommodative monetary policy environment in reality, this section examines the moderating effects of traditional monetary policy. Referring to relevant literature on the construction methods of traditional monetary policy (denoted as TMP), this section uses the M2 growth rate as a proxy for quantitative instruments and the market-oriented 7-day interbank offered rate as a proxy for price-based instruments (Gu & Bian, 2022; Huang & Guo, 2021). For the interbank offered rate, to match the annual frequency of the microdata in this paper, this section calculates the weighted average based on the implementation duration to obtain annual data for price-based monetary policy. It should be noted that to maintain consistency in the symbols of these variables, the inverse of the interbank offered rate is used as a measurement indicator, with both increasing representing an accommodative monetary policy environment.

Table 16. Results of moderating effects of traditional monetary policy.

	Quantitative				Price-based
	$/$		Macroprudential		$/$		Macroprudential
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
VARIABLES	$R W A R$	$Z_r$	$R W A R$	$Z_r$	$R W A R$	$Z_r$	$R W A R$	$Z_r$
$M P_{t} D_{i} T M P_{t}$	0.059**	0.017**			0.039**	0.010*
	(2.1924)	(2.0753)			(1.9804)	(1.9165)
$M P_{t} D_{i}$	0.976**	0.309**	1.019**	0.331**	0.981**	0.311**	1.015**	0.328**
	(2.4819)	(1.9935)	(2.4583)	(2.0197)	(2.5293)	(1.9846)	(2.5069)	(2.0186)
$M P_{t} D_{i} M P P_{t} T M P_{t}$			−0.124**	−0.056**			−0.074*	−0.038*
			(−2.3162)	(−2.1539)			(−1.8356)	(−1.9017)
$M P_{t} D_{i} M P P_{t}$			−0.292***	−0.168**			−0.322***	−0.171**
			(−2.6756)	(−2.4013)			(−2.6648)	(−2.3879)
$C o n s t a n t$	19.157	1.495**	22.462	1.698**	19.279	1.512**	22.217	1.629**
	(1.4761)	(1.9769)	(1.4971)	(1.9863)	(1.4837)	(1.9785)	(1.4581)	(1.9903)
$B a n k C o n t r o l s$	√	√	√	√	√	√	√	√
$M a c r o C o n t r o l s$	√	√	√	√	√	√	√	√
$I n d i v i d u a l F E$	√	√	√	√	√	√	√	√
$T i m e F E$	√	√	√	√	√	√	√	√
$O b s e r v a t i o n s$	787	787	787	787	787	787	787	787
$R - s q u a r e d$	0.173	0.314	0.175	0.316	0.173	0.314	0.174	0.316
$N u m b e r o f i d$	79	79	79	79	79	79	79	79

Note: * indicates significant at the 10% level, ** indicates significant at the 5% level, and *** indicates significant at the 1% level. The numbers in parentheses are t values.

Table 16 shows the test results of the moderating effects of traditional monetary policy. Among them, columns (1) to (4) show the moderating effects of quantitative monetary policy, with the first two columns showing regression results without macroprudential instruments, and the latter two columns incorporating macroprudential instruments. Similarly, columns (5) to (8) show the moderating effects of price-based monetary policy, arranged in the same way as the first four columns, which will not be repeated here. It can be seen that whether quantitative or price-based monetary policy, accommodative traditional policy instruments increase credit distortions overall, exacerbating the negative effects of structural monetary policy. After incorporating macroprudential instruments, their mitigating effect on structural monetary policy’s negative effects remains significant, while accommodative traditional monetary policy instruments can amplify the mitigating effect of macroprudential policy. Similarly, the effects of restrictive traditional monetary policy instruments take the opposite sign on this basis.

5. Conclusions and Policy Implications

Structural monetary policy, as an important instrument for central banks to regulate financial markets, directly or indirectly affects commercial banks’ asset quality, liquidity conditions, and credit strategies, ultimately influencing their risk-taking levels. Macroprudential instruments, through expectation guidance, demand suppression, and supply constraint mechanisms, effectively mitigate the risk spillover effects of structural monetary policy. The coordinated implementation of both enables central banks to attain structural-adjustment goals while maintaining economic and financial stability.

This paper thoroughly examines how structural monetary policy affects commercial banks’ risk-taking behaviors and how macroprudential instruments moderate these effects. Specifically, it focuses on the impact of the collateral expansion policy for lending facilities on commercial banks’ risk-taking and tests relevant hypotheses based on theoretical analysis. First, it constructs an identification strategy for the proportion of eligible collateral held by commercial banks using banking databases and public reports, and verifies the impact of the collateral expansion policy implemented in June 2018 on commercial banks’ risk-taking. Meanwhile, macroprudential instruments can effectively mitigate the negative effects of the collateral expansion policy.

Furthermore, this paper conducts an in-depth analysis of the transmission mechanism of structural monetary policy, verifying that the collateral expansion policy increases commercial banks’ risk-taking levels by raising targeted credit allocations to small and micro enterprises, green sectors, and agriculture-related sectors. At the same time, macroprudential policy constrains and optimizes relatively high-risk asset allocation behaviors, reducing the negative impact of the collateral expansion policy on commercial banks’ risk-taking. Notably, this paper verifies the robustness of the aforementioned conclusions by replacing risk-taking measures, macroprudential indicators, mechanism variables, identification strategies, placebo tests, and controlling for state capital injections and the COVID-19 pandemic.

Additionally, for different types of commercial banks, the collateral expansion policy and macroprudential policy demonstrate heterogeneous effects. Specifically, joint-stock and city commercial banks are more sensitive to policy changes, while large state-owned, rural, and foreign commercial banks remain relatively stable. Moreover, among various macroprudential policies, loan-based instruments have the most significant effect on structural monetary policy, followed by capital-based instruments, then liquidity-based instruments, and finally other types. Finally, this paper discusses the moderating effects of traditional monetary policy and finds that accommodative traditional monetary policy instruments exacerbate the risk spillover effects of structural monetary policy while amplifying the mitigating effects of macroprudential instruments.

Based on this, this paper offers the following policy recommendations:

First, to enhance the precision and safety of structural monetary policy, establish dynamic risk control mechanisms. For example, regularly conduct quantitative analysis of market liquidity and risk characteristics of emerging collateral, such as green bonds, dynamically adjust their risk conversion factors, and prevent commercial banks from excessively expanding risk exposure due to collateral expansion. Meanwhile, strengthen the countercyclical adjustment function of macroprudential instruments and implement differentiated regulation for commercial banks holding higher proportions of eligible collateral, such as dynamic provision coverage ratio requirements or leverage ratio restrictions, to curb irrational increases in their risk preferences.

Second, establish linkage mechanisms between targeted regulation policies and risk constraints, deeply integrating structural monetary policy incentives with macroprudential instruments. For instance, when providing medium-term lending facility incentives for commercial banks’ green loans, stipulate that green-loan non-performing ratios remain below industry benchmarks. For joint-stock and city commercial banks with high risk sensitivity, combine loan-based macroprudential instruments (e.g., loan growth ceilings) with capital-based instruments (e.g., countercyclical capital buffers). For large state-owned commercial banks, focus more on liquidity macroprudential instruments (e.g., liquidity coverage ratios) to form differentiated policy toolkits and achieve dual objectives of “precision irrigation” and “risk isolation”.

Finally, when implementing broad-based easing policies such as reserve requirement ratio cuts and interest rate reductions, proactively assess their superimposed effects with structural instruments. For example, when market interest rates fall beyond certain limits, trigger additional capital requirements for city commercial banks to prevent excessive expansion of high-risk collateral businesses using low-cost funds. Meanwhile, relying on the regulatory data platforms of central banks and financial supervision authorities, establish real-time monitoring systems for commercial banks’ risk-taking indices and implement necessary intervention measures for banks with abnormal fluctuations, thereby improving the “precision” of regulatory instruments while maintaining economic and financial stability.

It is worth noting that this paper is based on China’s localized dual-pillar regulatory framework and the domestic collateral expansion policy launched in 2018, and all empirical findings are derived from China’s unique institutional settings including banking ownership structure, monetary transmission rules and macroprudential supervision arrangements. Given substantial cross-country heterogeneity in central bank collateral frameworks, banking market structure and implementation details of macroprudential regulation across developed economies and other jurisdictions, the research conclusions have limited direct applicability for overseas banking systems. Cross-border comparative research based on multi-country panel data can be further explored in future studies to improve external validity.

NOTES

*The author is a postdoctoral researcher jointly trained by China Cinda Asset Management CO., Ltd. and the School of Applied Economics at Renmin University of China.

Conflicts of Interest

The author declares no conflicts of interest regarding the publication of this paper.

References

[1]	Adrian, T., & Liang, N. (2018). Monetary Policy, Financial Conditions, and Financial Stability. International Journal of Central Banking, 14, 73-131.
[2]	Albertazzi, U., Nobili, A., & Signoretti, F. M. (2021). The Bank Lending Channel of Conventional and Unconventional Monetary Policy. Journal of Money, Credit and Banking, 53, 261-299. [Google Scholar] [CrossRef]
[3]	Barth, J. R., Caprio, G., & Levine, R. (2013). Bank Regulation and Supervision in 180 Countries from 1999 to 2011. Journal of Financial Economic Policy, 5, 111-219. [Google Scholar] [CrossRef]
[4]	Bernanke, B. S., & Gertler, M. (1995). Inside the Black Box: The Credit Channel of Monetary Policy Transmission. Journal of Economic Perspectives, 9, 27-48. [Google Scholar] [CrossRef]
[5]	Borio, C., & Zabai, A. (2018). Unconventional Monetary Policies: A Re-Appraisal. In P. Conti-Brown, & R. M. Lastra (Eds.), Research Handbook on Central Banking (pp. 398-444). Edward Elgar Publishing. [Google Scholar] [CrossRef]
[6]	Borio, C., & Zhu, H. (2012). Capital Regulation, Risk-Taking and Monetary Policy: A Missing Link in the Transmission Mechanism? Journal of Financial Stability, 8, 236-251. [Google Scholar] [CrossRef]
[7]	Chen, G. J., Ding, S. J., Zhao, X. Q. (2024). Green Structural Monetary Policy, Policy Expectations and Green Economic Transition. Financial Research, No. 4, 1-19. (In Chinese)
[8]	Claessens, S., Ghosh, S. R., & Mihet, R. (2013). Macro-Prudential Policies to Mitigate Financial System Vulnerabilities. Journal of International Money and Finance, 39, 153-185. [Google Scholar] [CrossRef]
[9]	Dell’ariccia, G., Laeven, L., & Suarez, G. A. (2017). Bank Leverage and Monetary Policy’s Risk-Taking Channel: Evidence from the United States. The Journal of Finance, 72, 613-654. [Google Scholar] [CrossRef]
[10]	Deng, W., Song, M., & Chen, X. B. (2021). Do Lending Facilities Affect the Loan Maturity Structure of Commercial Banks? Finance & Trade Economics, 42, 83-96. (In Chinese)
[11]	Deng, X., Gan, S. T., & Liu, C. Z. (2023). Research on the Green Effect of MLF Collateral Expansion Policy: Based on the Perspective of Green Bond Financing Cost. International Finance Research, No. 10, 74-84. (In Chinese)
[12]	Fang, Y., Zhao, S. M., & Xie, X. W. (2012). Analysis of Monetary Policy and Bank Risk-Taking: Also on the Coordination of Monetary Policy and Macro-Prudential Policy. Management World, No. 11, 9-19+56+187. (In Chinese)
[13]	Gu, H. F., & Bian, Y. C. (2022). The “Dual-Pillar” Policy Framework, Cross-Border Capital Flows and Banking Systemic Risk. World Economy Studies, No. 11, 72-88+136. (In Chinese)
[14]	Hu, R. C., & Zhang, W. Q. (2016). Will the Implementation of Green Credit Affect the Profitability of Commercial Banks? Financial Regulation Research, No. 7, 92-110. (In Chinese)
[15]	Huang, Z., & Guo, Y. (2021). Central Bank Collateral Framework, Bond Credit Spreads and Corporate Financing Cost. Economic Research Journal, 56, 105-121. (In Chinese)
[16]	Jiang, S. X., & Chen, Y. C. (2012). Monetary Policy, Bank Capital and Risk-Taking. Financial Research, No. 4, 1-16. (In Chinese)
[17]	Jiménez, G., Ongena, S., Peydró, J., & Saurina, J. (2012). Credit Supply and Monetary Policy: Identifying the Bank Balance-Sheet Channel with Loan Applications. American Economic Review, 102, 2301-2326. [Google Scholar] [CrossRef]
[18]	Jin, C. X., & Jiang, X. (2020). Research on Policy Coordination and Its Applicability under the Dual-Pillar Regulation Mode of Monetary Policy and Macro-Prudential Policy. Collected Essays on Finance and Economics, No. 5, 43-53. (In Chinese)
[19]	Jin, P. H., Zhang, X., & Gao, F. (2014). Excessive Bank Risk-Taking and the Coordination between Monetary Policy and Counter-Cyclical Capital Adjustment. Economic Research Journal, No. 6, 73-85. (In Chinese)
[20]	Kashyap, A. K., & Stein, J. C. (2000). What Do a Million Observations on Banks Say about the Transmission of Monetary Policy? American Economic Review, 90, 407-428. [Google Scholar] [CrossRef]
[21]	Laeven, L., & Levine, R. (2008). Bank Governance, Regulation and Risk Taking. Journal of Financial Economics, 93, 259-275. [Google Scholar] [CrossRef]
[22]	Liu, W. (2016). Research on the Effectiveness and Optimization of Structural Monetary Policy Based on International Comparison. Journal of Financial Development Research, No. 9, 37-44. (In Chinese)
[23]	Ma, Y., & Chen, Y. L. (2013). Coordination and Combination of Macro-Prudential Policies: A Simulation Analysis Based on China. Financial Research, No. 8, 57-69. (In Chinese)
[24]	Ma, Y., & Yao, C. (2021). Effects of Monetary Policy and Macro-Prudential Policy under the Dual-Pillar Framework: Evidence from the Perspective of Bank Risk-Taking. Management World, 37, 51-69+3. (In Chinese)
[25]	Qi, J. Y., & Liu, Y. (2021). The Impact of “Dual-Pillar” Regulation on Commercial Banks’ Risk-Taking. International Finance Research, No. 9, 55-64. (In Chinese)
[26]	Qian, S. T., & Wu, W. H. (2020). Can Targeted Reserve Requirement Cuts Effectively Alleviate the Financing Difficulties of SMEs? Evidence from a Quasi-Natural Experiment Based on Bank-Level Micro-Data. Zhejiang Social Sciences, No. 11, 14-22+155. (In Chinese)
[27]	Schoemaker, P. J. H. (1993). Determinants of Risk-Taking: Behavioral and Economic Views. Journal of Risk and Uncertainty, 6, 49-73. [Google Scholar] [CrossRef]
[28]	Shin, H. S. (2016). Macroprudential Tools, Their Limits, and Their Connection with Monetary Policy. In O. Blanchard et al. (Eds.), Progress and Confusion (pp. 99-106). The MIT Press. [Google Scholar] [CrossRef]
[29]	Wang, Y. Q., & Wu, X. (2019). How Does China’s Innovative Monetary Policy Work: The Collateral Channel. Economic Research Journal, No. 12, 86-101. (In Chinese)
[30]	Yang, H. L., Fang, Y., Li, X. D. et al. (2017). Research on the Effectiveness of Macro-Prudential Policy in BRICS Countries. Macroeconomics Research, No. 1, 163-175. (In Chinese)
[31]	Zhang, Q., Qiao, Y. F., & Zhang, B. (2013). Does the Bank Risk-Taking Channel of Monetary Policy Exist in China? Financial Research, No. 8, 84-97. (In Chinese)
[32]	Zhang, X. L., & He, D. X. (2012). Monetary Policy Stance and Bank Risk-Taking: An Empirical Study on China’s Banking Industry (2000-2010). Economic Research Journal, No. 5, 31-44. (In Chinese)
[33]	Zhou, J., & Tao, S. G. (2019). The Impact of Structural Monetary Policy on the Efficiency of Chinese Commercial Banks: A Study Based on the Bank Risk-Taking Channel. China Economic Issues, No. 3, 25-39. (In Chinese)

	[email protected]
	+86 18163351462 (WhatsApp)
	1655362766
	SCIRP WeChat

Journals Menu

Home

About SCIRP

Service

Policies