Lender of Last Resort and Financial Systemic Risks in Times of Economic Stability: Evidence from 55 Countries

Abstract

As a cornerstone of the modern financial safety net, the Lender of Last Resort (LOLR) is essential in mitigating liquidity crises and containing financial contagion. However, during periods of economic stability, risk-taking incentives in the banking sector may undermine its effectiveness. Using quarterly panel data from 55 countries over the period 2010–2023, this study employs a two-way fixed effects model to assess the impact of LOLR support on systemic financial risk and its transmission mechanisms. We find that LOLR support significantly increases systemic risk during stable economic periods. Mechanism analysis indicates that this effect is channeled through the erosion of bank asset liquidity, expansion of financial leverage, and deterioration in asset quality. Moreover, the adverse impact is more pronounced in emerging economies, bank-dominated financial systems, countries with low capital adequacy ratios, underdeveloped regulatory frameworks, and lower levels of digital technology adoption. This study provides cross-country evidence on the potential negative consequences of central bank rescue functions during calm periods and offers important policy insights for optimizing the LOLR framework and building a more resilient financial safety net.

1. Introduction

As one of the three pillars of the financial safety net, the Lender of Last Resort (LOLR) aims to resolve financial institution crises and curb risk contagion through central bank liquidity support. Theoretically, the LOLR can provide emergency liquidity when financial institutions face adverse shocks and market financing fails, thereby containing the spread of risk (Bagehot, 1873). However, after the 2008 global financial crisis, the use of LOLR facilities has increasingly become a routine source of liquidity in practice (Nelson, 2014; Dobler et al., 2016). This frequent reliance on LOLR may undermine its policy effectiveness, foster moral hazard among financial institutions, and even increase their risk-taking behavior (Fecht & Weber, 2023), posing a potential threat to financial stability. Therefore, further clarifying the actual impact and transmission channels of the LOLR on systemic financial risk is of significant research importance for improving the LOLR rescue framework and maintaining financial stability.

The existing literature has extensively discussed the role of LOLR in maintaining financial stability, primarily focusing on its theoretical design and practical challenges. Some studies affirm the LOLR’s critical role in resolving liquidity crises among financial institutions and preventing risk contagion (Summers, 1991; Fischer, 1999). They argue that by imposing punitive interest rates and requiring eligible collateral, the LOLR can provide liquidity support while curbing moral hazard, thereby breaking the vicious cycle of “market panic—sell-off—liquidity depletion” (Tucker, 2014) and ultimately safeguarding financial stability. However, as LOLR has been applied in practice, numerous studies have also highlighted its negative implications. During actual crisis responses, central banks struggle to accurately distinguish between institutions experiencing temporary liquidity shortages and those facing solvency crises within short timeframes. This difficulty makes it challenging to precisely determine which banks should receive loans (Freixas et al., 2000), potentially leading to liquidity support being extended to insolvent institutions (Goodhart, 1987). This broadening of the rescue scope may encourage “zombie institutions” to continue operating, thereby delaying risk resolution and exacerbating the cyclical accumulation of financial risks. Simultaneously, the act of LOLR rescue itself may generate risks. On one hand, the externalization of losses may increase financial institutions’ willingness to take on risks (Dam & Koetter, 2012). On the other hand, rescuing “too big to fail” institutions may distort market incentives and reinforce the “too big to fail” problem (Stiglitz, 2010). Particularly when non-bank financial institutions also anticipate central bank liquidity support, moral hazard may further spread beyond the scope of regulatory oversight (Avernas et al., 2020). Following the financial crisis, the operational framework of the LOLR has gradually expanded, with rescue methods evolving from traditional credit support to acting as a market maker of last resort. While such operations help mitigate market failures, they may also foster a liquidity illusion, leading financial institutions to become overly reliant on central bank risk guarantees (Dam & Koetter, 2012) and thereby increasing financial vulnerability.

Existing research has primarily focused on the effectiveness of the LOLR during financial crises, reaching broad consensus on two key points. First, during financial crises, adequate LOLR support can mitigate the risk of bankruptcy for financially distressed institutions facing liquidity challenges. Second, if the LOLR fails to accurately identify institutions experiencing temporary liquidity difficulties, overly broad liquidity assistance may foster dependence on central banks for risk resolution. However, existing research has paid less attention to the fact that after the 2008 financial crisis, many countries entered periods of economic stabilization and recovery. Their central banks continued to provide extensive liquidity support, employing tools and methods similar to those used during the financial crisis. This may lead to the broadening of the LOLR, potentially weakening its effectiveness in maintaining financial stability during periods of economic calm. Specifically, existing research lacks a systematic global perspective examining the impact of the LOLR on financial stability during stable economic periods and its potential transmission channels. To address this gap, this paper aims to answer the following questions through empirical research: What is the actual impact of the LOLR on systemic financial risks under stable economic conditions? Through which channels does this effect operate? And how does this impact vary across countries based on their economic development and institutional characteristics?

To address the research questions, this study examines the impact of the LOLR on systemic financial risk and its underlying mechanisms using quarterly panel data from 55 countries spanning the period 2010–2023. The findings reveal that: (1) During tranquil economic periods, the LOLR contributes to an increase in systemic financial risk across countries. (2) Further analysis indicates that the LOLR elevates systemic risk primarily through the erosion of bank asset liquidity, expansion of financial leverage, and deterioration in asset quality. (3) Heterogeneity analysis shows that the risk-augmenting effect of the LOLR is more pronounced in emerging and developing economies, countries with bank-dominated financial systems, those with lower capital adequacy, nations characterized by weaker regulatory quality, and those with lower levels of digital technology development.

The marginal contributions of this paper are threefold. First, it elucidates how LOLR exacerbates and accumulates systemic risk during stable economic conditions. Prior research on the real effects of LOLR has largely focused on specific crisis episodes, such as the global financial crisis (Jeffers, 2010) and the European sovereign debt crisis (Drechsler et al., 2016). While studies on crisis-period liquidity assistance remain vital, financial institutions tend to exhibit stronger profit-seeking incentives during stable times. The financial risks accumulated in such periods are more likely to materialize during downturns, making their impact on systemic risk non-negligible. This study thus complements the literature by examining the consequences of LOLR interventions in non-crisis times, offering relevant insights for systemic risk prevention in the post-crisis era.

Second, this paper analyzes the asset allocation patterns of financial institutions after receiving liquidity support during periods of economic stability. Banks are inherently risk-bearing institutions characterized by the privatization of profits and the socialization of risks. This inherent trait makes them prone to developing expectations of bailout dependency. To address this, we identify three specific mechanisms through which LOLR aggravates systemic risk. This provides empirical evidence on the transmission channels of LOLR-induced financial risk.

Third, this study examines how the impact of LOLR on systemic risk varies across countries with different levels of economic development, financial structures, capital adequacy, regulatory quality, and digital technology adoption. Existing research has largely overlooked the context-dependent nature of LOLR’s influence on systemic risk. This paper supplements LOLR research on systemic risk environment differences using national samples. It clarifies the impact of LOLR on systemic risk across different financial systems, expanding the breadth and depth of LOLR studies. These insights not only contribute to refining the crisis resolution framework but also offer important evidence for tailoring financial risk regulations across diverse national contexts.

2. Theoretical Analysis and Hypothesis

2.1. LOLR and Financial Systemic Risk

The original role of the LOLR was to provide emergency credit support to banks experiencing temporary liquidity shortages but possessing repayment capacity, thereby preventing the spread of crises (Bagehot, 1873). However, in the implementation of LOLR policies, even when central banks extend liquidity support at punitive interest rates to financial institutions facing temporary liquidity crunches, such measures may still exacerbate risk-taking behavior among these institutions.

First, from the perspective of financial institutions’ risk decision-making, once banks anticipate central bank liquidity support, their risk appetite systematically increases, leading them to favor investments in high-risk assets (Drechsler et al., 2016; Fecht & Weber, 2023). Given the high interconnectedness of interbank assets, this rise in individual risk appetite readily evolves into the accumulation of systemic risk (Acemoglu et al., 2015). Second, regarding the information and identification challenges faced by central banks, they struggle to swiftly and accurately distinguish between banks facing liquidity shortages and those that have already lost solvency during crises (Freixas et al., 2000). This “identification problem” often leads to overly broad bailouts, allowing some “zombie banks” that should have exited the market to continue operating, thereby sowing seeds of future risk (Goodhart, 1987). Finally, under government intervention, the LOLR may deviate from its goal of financial stability. It can become instead a tool for serving government debt financing or maintaining regulatory credibility (Drechsler et al., 2016). For instance, during sovereign debt crises, governments may encourage banks to purchase sovereign bonds to alleviate debt pressures (Buiter & Rahbari, 2012). Alternatively, public funds may be used to conceal bank losses for political reputational reasons (Boot & Thakor, 1993; Mishkin, 2000), actions that further amplify the banking system’s risk exposure.

Therefore, we acknowledge the risk-absorbing effect of LOLR during the financial crisis. However, following the financial crisis, central banks in multiple countries expanded their balance sheets to provide more ample liquidity support to the market (Moessner & Allen, 2010). This expansionary behavior may lead financial institutions to rely on central banks to mitigate their operational risks, thereby intensifying banks’ risk-taking propensity due to the externalization of losses (Lerrick & Meltzer, 2003; Dam & Koetter, 2012). During stable economic periods, central banks’ LOLR liquidity support may trigger behavioral risk preferences among financial institutions, ultimately leading to the unintended accumulation of systemic risk.

Therefore, this paper proposes:

Hypothesis 1.

During periods of economic stability, LOLR support exacerbates the accumulation of systemic financial risks.

2.2. Mechanisms of LOLR Support on Financial Systemic Risk

2.2.1. LOLR Support, Liquidity Weakening Mechanism and Financial Systemic Risk

The actual impact of LOLR support may depend on how financial institutions allocate assets after receiving liquidity (Drechsler et al., 2016). If the funds are used for high-risk investments or regulatory arbitrage, the LOLR could ultimately undermine financial stability. Excessively broad liquidity assistance can foster a “liquidity illusion” in the market during normal periods (Moe, 2012). When banks anticipate a central bank bailout during a liquidity crisis, they are inclined to adopt a suboptimal liquidity management strategy (Ratnovski, 2009) by augmenting their allocation to higher-yielding but less liquid long-term assets (Delis & Kouretas, 2011). In a fully competitive environment, projects with higher expected returns, shorter payback periods, higher liquidity, and lower risk have long been fully competed for and occupied, resulting in most less-competitive banks possessing only a limited number of residual investments, which typically exhibit common traits—illiquidity or risk-return imbalance. The uniformity in bank behavior has led to an exacerbation of the asset-liability maturity mismatch within the banking system, and the structural disparity between short-term liabilities and long-term assets may heighten liquidity vulnerability (Silva, 2019). In nations with punitive interest rates, banks will incur the punitive rate irrespective of their portfolio allocation, leading them to prefer long-term assets to secure a profitable return. In addition, nations that provide special interest rates for bailouts often stimulate bank arbitrage activities. Specifically, acquiring inexpensive rescue capital from the central bank to invest in long-term assets for enhanced returns (Sim, 2024). The aforementioned allocation practices of banks exacerbate the accumulation of risk inside the financial system.

Therefore, this paper proposes:

Hypothesis 2.

LOLR increases financial systemic risk by weakening asset liquidity mechanisms in the banking system.

2.2.2. LOLR Support, Leverage Expansion Mechanism and Financial Systemic Risk

The transmission of LOLR support to financial systemic risk may also be evident in the apparent expansion of bank leverage. The LOLR can stabilize market confidence in the short term. In the long term, LOLR support indicates the central bank’s assurance to the market, which diminishes the market’s inclination to supervise banks due to the central bank’s endorsement (Brandao-Marques et al., 2020), thereby undermining the market’s disciplinary role in bank supervision (Rochet & Tirole, 1996). With LOLR backing and declining debt financing costs, banks will opt to rapidly increase their asset and liability sizes, resulting in much greater leverage. Furthermore, the anticipation of central bank support may compel banks to negotiate regulatory concessions owing to their “too big to fail” status, resulting in heightened asset bubbles via leverage amplification during economic upswings and an escalation of procyclical risk within the financial system due to deleveraging practices that provoke asset sell-offs and liquidity shortages during downturns. This irrational escalation in bank leverage heightens individual risk exposure and establishes a chain of risk contagion via asset-liability connections, ultimately exacerbating financial systemic risk (Cincinelli et al., 2024). This mechanism may be influenced by the capital adequacy of the financial system and the quality of regulation (Drechsler et al., 2016). In nations with inadequate capital reserves and substandard regulation, banks often augment their financial leverage to achieve elevated profits necessary for capital replenishment to satisfy regulatory mandates. Concurrently, the absence of stringent oversight leads banks to perceive that heightened risky leverage will evade regulatory scrutiny, thereby amplifying risk exposure within the banking system.

Therefore, this paper proposes:

Hypothesis 3.

LOLR expands via banking system leverage, thereby elevating financial systemic risk.

2.2.3. LOLR Support, Asset Quality Deterioration Mechanism and Financial Systemic Risk

The consequences of banks’ allocation of long-maturity, high-risk projects and leveraging operations are directly manifested in asset quality deterioration. First, from a banking standpoint, LOLR support reduces the cost of risk-taking for banks, hence incentivizing the relaxation of credit assessment requirements, which results in a decline in asset quality. Furthermore, owing to insufficient regulatory penetration in certain nations, banks have excessively engaged in off-balance-sheet businesses, such as shadow banking, in pursuit of high-risk returns, resulting in diminished asset quality within the banking system. Second, from a regulatory standpoint, central banks tend to exhibit a lenient approach towards problematic bank assets to mitigate risk exposures to prevent systemic panic. Especially in practice, the central bank finds it challenging to precisely differentiate between solvent and insolvent institutions within a limited timeframe, complicating decisions regarding which banks should receive loans (Freixas et al., 2000), necessitating the implementation of broad liquidity support. This may contribute to maintaining ostensible operations of institutions on the verge of bankruptcy due to ongoing liquidity bailouts (Goodhart, 1987), resulting in increased financial risk accumulation (Bindseil & Jablecki, 2023). During economic downturns or when the financial system faces external shocks, the prolonged accumulation of non-performing assets will be revealed in a concentrated manner, resulting in a domino effect of risk contagion. The decline in asset quality of individual banks disseminates through guarantees and credit-linked channels, such as counterparties, when risk surpasses a critical threshold, and it can precipitate a recession in the real economy due to a credit crunch, thereby intensifying systemic fragility (Acharya & Yorulmazer, 2007; Farhi & Tirole, 2012; Freixas et al., 2004).

Therefore, this paper proposes:

Hypothesis 4.

LOLR exacerbates financial systemic risk via the mechanism of asset quality deterioration in the banking system.

3. Research Design

3.1. Econometric Model Setting

This study examines the impact of LOLR on financial systemic risk. The fixed effects model is constructed as follows:

$$SY{S}_{i,t}={\beta }_0+{\beta }_{1}LOL{R}_{i,t-1}+{\beta }_2{X}_{i,t-1}+{\mu }_i+{\lambda }_t+{\epsilon }_{i,t}$$

(1)

Among them, $SY{S}_{i,t}$ denotes the financial systemic risk of country $i$ in period $t$. $LOLR$ is the ratio of total central bank claims on the banking system to the total assets of the banking system, which measures the intensity of the LOLR bailout of the banking system as a whole. The parameter ${\beta }_1$ captures the effect of LOLR on financial systemic risk. $X_{i,t-1}$ is a control variable, ${\mu }_i$ denotes country fixed effects, ${\lambda }_t$ denotes time fixed effects, and ${\epsilon }_{i,t}$ denotes a random disturbance term. The paper mitigates potential endogeneity problems by lagging all explanatory variables by one period and clustering the regression standard errors at the country level.

3.2. Variable Selection

1. Systemic risk ($SYS$). We selected SRISK as the metric for systemic risk primarily based on this study’s core requirements for forward-looking and real-time indicators. This study aims to explore the potential systemic risks that LOLR may generate within the financial system, rather than merely measuring realized risks. Consequently, a forward-looking risk metric is crucial. Among widely recognized systemic risk measures such as MES, SES, and ΔCoVaR, SRISK stands out as a forward-looking, market-data-driven expected indicator (Brownlees & Engle, 2017). By simulating future market extreme scenarios to calculate financial institutions’ expected capital shortfalls, it directly captures potential systemic vulnerabilities induced by the LOLR. This key characteristic makes SRISK more suitable for this study’s analysis compared to other risk measures.

Specifically, the indicator is derived by summing the SRISK values of financial institutions with risk contribution (SRISK%) exceeding 0 (Brownlees & Engle, 2017). The indicator calculates the expected capital shortfall of a financial institution in a potential systemic crisis, taking into account factors such as financial institutions’ leverage, long-term marginal losses, and size, and accurately measures systemic risk in the event of a sustained economic downturn, while providing good predictability for the macroeconomy (Yang et al., 2022). The indicator consists of two core computational steps. The first is to measure the long-term marginal expected loss LRMES based on a dynamic conditional beta model simulated with market extreme scenarios. Specifically, the regression equation is defined by mixing a time-varying beta model and a constant beta model: $r_{i,t}=\left({\Phi }_1+{\Phi }_2{\beta }_{i,t}\right)r_{m,t}+{\mu }_t$, this equation is set up to ignore the noise caused by the covariance estimation error. To provide a better fit to the available data, an assumption of heteroskedasticity is made, so that the equation can be deformed as follows: $r_{i,t}=\left({\Phi }_1+{\Phi }_2{\beta }_{i,t}\right)r_{m,t}+\sqrt{h_{i,t}}{\xi }_{i,t}$. Following the combination of GJR-GARCH with dynamic correlation DCC to capture asymmetric volatility as well as time-varying correlation, Bootstrap sampling is utilized to generate return paths for the next six months. Using the U.S. as an example, the potential loss in the market value of the institution’s stock is calculated for the next six months when the decline in the S&P 500 index breaks the crisis threshold. The formula is defined as: $LRMES=1-exp\left(log\left(1-d\right)*beta\right)$, where d denotes the crisis threshold for the market to fall in six months (default is 40%) and beta is the beta coefficient of the institution’s stock. In the second step, a dynamic assessment of capital reserves needs to be calculated. After measuring the expected stock price loss during the crisis, the minimum capital reserves required for a financial institution to withstand a financial crisis of the same magnitude need to be calculated by combining the current market value of equity with on-balance sheet liabilities. The formula for this calculation is: $SRISK=k\cdot D\mathrm{e}bt-(1-k)Equity\cdot (1-Lrmes)$, where k is the level of sound capital adequacy, Debt is the book value of the financial institution’s liabilities, Equity is the market value of the financial institution’s stock, and Lrmes is the long-run expected marginal loss. A positive value of SRISK represents that the financial institution is systemically risky. Larger values represent higher systemic risk of the financial institution. In addition, due to the large magnitude of this data, it is logarithmically normalized in this paper.

2. Explanatory variables (LOLR). The core explanatory variable in this paper is the intensity of implementation of the central bank’s LOLR function. According to Praet (2016), central banks primarily exercise their LOLR role through two approaches: the credit approach and the monetary approach. In practice, the instruments associated with these approaches (such as re-lending, standing lending facilities, and open market operations) are often used in combination. Therefore, the proxy variable for the intensity of LOLR support is the ratio of the total claims held by central banks on the banking system to the total assets of the banking system.

3. Control variables. Referring to (Ren et al., 2023; Svahn et al., 2017), this research selects control variables from two dimensions, the macroeconomic environment and the features of the financial system, to mitigate the influence of extraneous influences on financial systemic risk: (1) the size of the banking sector: Measured using the natural logarithm of the total assets of the banking sector across various countries. Large financial institutions may be expected to be “too big to fail” with higher risk contagion, so the impact of size on systemic risk needs to be controlled. (2) Capital Adequacy Ratio (CAR): Calculated as the ratio of total capital to total risk-weighted assets in the banking sector. The level of bank capital buffer will directly affect the risk-absorbing capacity, so this paper controls its impact on systemic risk. (3) Return on Assets (ROA): Calculated using the ratio of net profit to total assets in the banking sector to control for the impact of changes in bank profitability on risk appetite, and thus the potential transmission to systemic risk. (4) Provision Coverage Ratio (NPLC): Measured using the ratio of loan loss provisions to non-performing loan balances in the banking sector to control for the possibility that low provision coverage may exacerbate risk-loss spillovers in times of crisis. (5) GDP growth rate (G): controls for the impact of economic cycle fluctuations on the financial system’s vulnerability. (6) Inflation rate (CPI): controls the impact of dramatic price volatility factors on systemic risk. (7) M2 growth rate (M2): controls the impact of quantitative monetary policy on financial system risk.

3.3. Data Description

This study constructs an initial sample based on systemic risk data published by the V-Lab, primarily covering 80 economies worldwide. To ensure research rigor, systematic sample screening is required. First, the sample period was defined as the first quarter of 2010 to the fourth quarter of 2023, encompassing a relatively complete macro-financial cycle in the post-financial crisis era. Based on this, countries with severe data gaps in core explanatory variables such as the LOLR were excluded. Furthermore, samples with substantial missing values in other key variables were also eliminated. Following this screening process, the final sample comprised 55 countries, totaling 3028 observations. Financial systemic risk data was sourced from the V-Lab website. Data for core explanatory variables, control variables, and mechanism variables originated from the CEIC database, World Bank Open Data, and IMF Data, among others. Considering the potential interference of outliers on research findings, all continuous variables underwent trimming at the 1% level. Table 1 presents specific descriptive information.

Table 1. Variables and descriptive statistics.

Variables	Variable Definition	N	Mean	St	Min	Max
SYS	Systemic Risk	3028	9.4719	4.7637	0	16.0753
LOLR	LOLR Support	3028	0.7727	5.6673	0	48.3781
SIZE	Asset Size	3028	15.6111	2.7695	9.7439	23.0474
CAR	Capital Adequacy Ratio	3028	16.6866	3.2750	9.8345	26.4048
ROA	Return on Assets	3028	1.1690	0.9120	−1.5188	4.4804
NPLC	Provision Coverage Ratio	3028	65.8200	40.8543	16.2985	213.8296
GDP	GDP Growth Rate	3028	2.8930	4.2298	−11.5137	16.7793
M2	M2 Growth Rate	3028	8.6355	7.8159	−9.2517	44.9669
CPI	Inflation Rate	3028	4.2560	5.9378	−1.5	40.5423
NPL	Non-Performing Loan Ratio	3028	5.0818	6.4494	0.3426	37.9307
LR	Liquidity Ratio	3028	58.2670	41.9502	15.7108	240.4953
LEV	Leverage Ratio	3028	2.6888	1.3189	0	6.9029

The sample selected for this study encompasses both developed economies (such as the United States, Germany, and Japan) and emerging and developing economies (such as China and India), covering countries at different stages of economic development. As shown in Table 2, the selected sample countries are highly representative.

Table 2. Sample countries.

Advanced Economies		Emerging Market and Developing Economies
Austria	Luxembourg	Argentina	Mexico
Australia	Malta	Bangladesh	Morocco
Belgium	Netherlands	Barbados	Nigeria
Canada	Norway	Brazil	Pakistan
Cyprus	Portugal	Chile	Panama
Denmark	Singapore	China	Philippines
Finland	Slovenia	Colombia	Poland
France	South Korea	Egypt	Russian Federation
Germany	Spain	Hungary	South Africa
Greece	Sweden	India	Sri Lanka
Ireland	Switzerland	Indonesia	Thailand
Israel	United Kingdom	Kazakhstan	Turkey
Italy	United States	Kuwait	United Arab Emirates
Japan		Malaysia	Vietnam

4. Result Analysis

4.1. Benchmark Regression Analysis

Table 3 presents the regression findings of LOLR support on financial systemic risk, as derived from model (1). This work clusters the standard errors at the country level to mitigate the impact of inter-country correlation on the research outcomes. The data in Table 3 indicate that the regression coefficients of LOLR support on SRISK across nations are considerably positive at the 1% level. This indicates that LOLR assistance exacerbates the vulnerability of the financial system and elevates systemic financial risk. Consequently, Hypothesis 1 is evaluated.

Table 3. Benchmark regression analysis.

	(1) SYS	(2) SYS	(3) SYS	(4) SYS
LOLR	0.0829 ***	0.0480 ***	0.0733 ***	0.0391 ***
	(39.6184)	(5.8644)	(20.7991)	(5.2916)
SIZE			0.1135	0.0831
			(1.6591)	(1.1116)
CAR			0.0841	−0.1846 *
			(1.0481)	(−1.9709)
ROA			−1.0888 ***	−0.9677 ***
			(−4.3908)	(−3.9282)
NPLC			−0.0087	−0.0162 **
			(−0.7730)	(−2.0760)
GDP			−0.0646 ***	−0.0363
			(−3.0310)	(−1.1018)
CPI			0.0921 ***	0.0413
			(2.6773)	(1.1397)
M2			−0.0588 **	−0.0615 ***
			(−2.6336)	(−2.9446)
C	9.4399 ***	8.3361 ***	9.3813 ***	12.9358 ***
	(5856.4389)	(15.0727)	(5.9054)	(7.3329)
N	2973	2973	2973	2973
National FE	YES	YES	YES	YES
Time FE	NO	YES	NO	YES
Control	NO	NO	YES	YES
R2	0.0024	0.0864	0.1130	0.2138

Note: ***, **, and * indicate significance at the 1%, 5%, and 10% levels, respectively. The t-values are in parentheses.

This empirical finding confirms the theoretical analysis presented earlier. The banking system tends to distort LOLR liquidity support into a form of “quasi-explicit bailout implicit in the system” (Drechsler et al., 2016; Fecht & Weber, 2023), fostering persistent expectations of central bank bailouts. Under such expectations, banks face distorted incentives characterized by “privatizing profits and socializing losses.” They reap upside investment returns while externalizing downside risks through anticipated bailouts, thereby strengthening their motivation to pursue high-risk, high-return projects. This systemic increase in risk appetite manifests as aggressive tendencies in banks’ asset allocation behavior. On the asset side, banks increase allocations to high-risk, long-duration assets. Coupled with the strong interconnectedness among banks stemming from holding similar assets, individual banks’ risky behavior not only undermines their own stability but also rapidly propagates through tight network ties during external shocks, ultimately amplifying systemic risk across the entire financial system (Fang & Liu, 2023).

4.2. Robustness and Endogeneity Test

(1) Replacement of the explained variables. This work reconstructs the systemic risk SRISK index hazard thresholds in numerous scenarios to avoid biased financial systemic risk estimation due to hazard threshold setting deviation. The original computational model’s 40% market drop thresholds are changed to 50% and 60%, recalculating each country’s financial system’s SRISK values. The new SRISK_50 and SRISK_60 depict predicted capital deficits under severe scenarios of a 50% and 60% market drop in six months. After replacing the explained variables, LOLR’s regression coefficients match the benchmark results in columns (1)–(2) of Table 4, showing this paper’s robustness.

Table 4. Robustness and endogeneity test.

	(1) SYS	(2) SYS	(3) SYS	(4) SYS
	(50%)	(60%)	DIF-GMM	SYS-GMM
LOLR	0.0575 ***	0.0486 ***	0.1557 *	0.1522 **
	(7.1159)	(6.7043)	(1.7473)	(2.3774)
SYSt-1			0.5853 ***	0.5708 ***
			(11.7490)	(10.8142)
SYSt-2			0.1955 ***	0.1858 ***
			(3.6429)	(3.4740)
AR (1)-P			0.000	0.000
AR (2)-P			0.657	0.722
Hansen-P			0.284	0.334
Control	YES	YES	YES	YES
National FE	YES	YES	YES	YES
Time FE	YES	YES	YES	YES
R2	0.1721	0.1709
N	2973	2973	2863	2918

Note: ***, **, and * indicate significance at the 1%, 5%, and 10% levels, respectively. The t-values are in parentheses.

(2) Re-estimation based on dynamic GMM. This work employs dynamic GMM for robustness testing to alleviate the dynamic panel bias associated with the estimate of the fixed-effects model. We incorporate lag 1 and lag 2 of the explained variables (systemic risk) as explanatory variables in the benchmark model. The tests utilize DIF-GMM and SYS-GMM, while adjusting for robust standard errors to address the heteroskedasticity of the error terms in the pertinent dimensions. The estimation results are presented in Table 4 columns (3)–(4), with p-values for Hansen’s test and AR(2) over 0.1, suggesting that the model is adequately specified. The lack of second-order autocorrelation in the disturbance term and the absence of over-identification of instrumental variables have been confirmed. The primary explanatory variable in the LOLR regression results aligns with the benchmark model, hence reinforcing the credibility of the empirical findings.

(3) Excluding anomalous time samples. During the COVID-19 pandemic, central banks predominantly employed unconventional monetary policy instruments to intervene in the economy, resulting in the interruption of the LOLR function and traditional economic cycle risk mitigation. This study re-estimates the data by removing the two-year sample of nations from 2020 to 2021 to ensure the robustness of the conclusions. The results shown in Table 5 column (1) demonstrate that the influence of LOLR support on systemic risk remains evident throughout periods of economic stability, underscoring the generalizability of this paper’s conclusions.

Table 5. Robustness and endogeneity test.

	(1)	(2)	(3)	(4)	(5)	(6)
	SYS	SYS	SYS	SYS	SYS	SYS
LOLR	0.0261 ***	0.0356 ***	0.0350 ***	0.0403 ***	0.0383 ***	0.0234 **
	(3.8658)	(4.8665)	(4.8233)	(5.4522)	(5.2095)	(2.3613)
IMR				−0.4554
				(−0.6860)
Control	YES	YES	YES	YES	YES	YES
National FE	YES	YES	YES	YES	YES	YES
Time FE	YES	YES	YES	YES	YES	YES
N	2478	3022	2988	2973	2973	2973
R2	0.1853	0.2252	0.2242	0.2141	0.2175	0.2349

Note: *** and ** indicate significance at the 1% and 5% levels, respectively. The t-values are in parentheses.

(4) Propensity score matching (PSM). This research uses PSM to examine the impact of LOLR support, thereby addressing the potential issue of sample selection bias. This study investigates whether the magnitude of central banks’ LOLR assistance to the banking sector exceeds the median, hence categorizing the sample into a high support group (experimental group) and a low support group (control group). The sample is aligned using a Logistic model with a 1:1 closest neighbor matching method without replacement, and the regression outcomes of the matched sample are presented in Table 5 column (2). This research re-evaluates the matched samples after trimming 1% of the data to mitigate the influence of outliers, with the results presented in Table 4 column (3). The investigation indicates that the LOLR regression coefficients remain significantly positive post-matching, thereby demonstrating the robustness of the propensity score-based matching regression.

(5) Heckman’s two-step methodology. This paper uses the Heckman two-stage model to rectify estimating bias arising from sample self-selection. It initially determines if the strength of a nation’s central bank’s LOLR support exceeds the median, formulates a probit regression, and computes the IMR. This paper adheres to the World Bank’s classification of countries into four income categories (high-income, upper-middle-income, lower-middle-income, and low-income) and selects the mean value of LOLR support for each economic income group within the same quarter as an exogenous impact variable. In the second stage, the IMR calculated in the first stage is incorporated into the model for estimation to address the self-selection issue. The estimation results are presented in Table 5 column (4). The LOLR coefficient estimates exhibit a substantial positive correlation, and the regression results are robust.

(6) Test for omitted variables. This research incorporates two essential categories of factors—national governance variables and digital technology development—into the benchmark regression model to avert the exclusion of critical control variables and mitigate estimation bias issues. National governance quality may affect financial system regulatory efficiency and risk-absorbing capacity. This article utilizes the World Bank Global Governance Index (WGI) to assess country governance efficacy. Digital technology may alter the pathways of risk contagion, and the ease of transactions may enhance the velocity of risk shocks. This study employs the Internet Penetration Rate Indicator to assess the extent of digital technology integration in each country, effectively reflecting the level of digital infrastructure and information technology utilization. Controlling for the exogenous shocks of the macro-institutional and technology variables captures LOLR’s potential impact on systemic risk. In Table 5, columns (5) and (6) show the estimation results, and the LOLR support coefficient is still significantly positive, confirming this paper’s empirical findings.

4.3. Mechanism Tests

4.3.1. Liquidity Weakening Mechanism

As a crucial element of the financial safety net, LOLR support seeks to prevent insolvency crises and risk contagion by offering emergency financing to institutions experiencing liquidity trouble. This research employs the banking system liquidity ratio as an explanatory variable to assess the actual liquidity position of banks following LOLR support. The estimation results are presented in Table 6 column (1). The regression coefficient of LOLR support on the banking system liquidity ratio is considerably negative at the 5% significance level. The findings indicate that banks receiving LOLR support weaken their liquidity management and choose to allocate long-maturity assets to enhance profitability in their daily operations. This unintended consequence occurs in both types of financial systems: in certain countries with penalized interest rates, banks typically allocate the supplementary liquidity bailout funds to long-maturity, high-yielding assets as a safeguard against elevated interest costs, resulting in further liquidity deterioration. Moreover, certain nations may contemplate employing preferential interest rates for their financial rescue. This would foster the development of bank arbitrage incentives, utilizing the central bank’s low-interest rate bailout funds to invest in long-term assets in search of greater profits. When the central bank constricts liquidity, the danger of maturity mismatch will swiftly proliferate, hence intensifying systemic risk. Consequently, Hypothesis 2 is confirmed.

Table 6. Mechanism test.

	(1)	(2)	(3)
	LR	LEV	NPL
LOLR	−0.1189 **	0.0208 ***	0.0560 ***
	(−2.0148)	(8.0263)	(7.6203)
Control	YES	YES	YES
National FE	YES	YES	YES
Time FE	YES	YES	YES
R2	0.0243	0.1817	0.1490
N	2973	2973	2973

Note: *** and ** indicate significance at the 1% and 5% levels, respectively. The t-values are in parentheses.

4.3.2. Leverage Expansion Mechanism

While LOLR support eases banks’ short-term repayment pressure, it also weakens the constraints of market supervision on their risky behaviors, leading to banks’ leveraging behaviors. To verify the actual impact of LOLR support on financial leverage, this paper takes the leverage ratio as an explanatory variable to test the mechanism. The estimation results are shown in Table 6 column (2); the regression coefficient of LOLR support on financial system leverage is significantly positive at the 1% level, indicating that LOLR support increases the level of financial system leverage. The economic reason for this is that, on the one hand, LOLR support weakens the market’s initiative to monitor financial risks, and this weakened constraint mechanism drives the willingness of banks and other institutions to increase leverage. On the other hand, financial institutions may ignore the long-term risks of leverage accumulation due to the reduced cost of risk-taking. Highly leveraged operations amplify the vulnerability to asset price volatility, and in the event of a negative market shock, financial systemic risk will subsequently intensify. Thus, Hypothesis 3 is validated.

4.3.3. Asset Quality Deterioration Mechanism

As previously stated, LOLR support amplifies the risk-taking tendencies of the banking system in the allocation of long-term assets and the augmentation of leverage, potentially resulting in heightened asset losses. This article employs the non-performing loan ratio of the banking sector to assess asset quality for the mechanism test. The estimation findings are presented in Table 6 column (3), indicating that the regression coefficient of LOLR support on the NPL ratio of the banking system is notably positive at the 1% level. The findings suggest that LOLR support leads to a decline in the banking sector’s asset quality. LOLR support diminishes banks’ stringent risk assessment standards, resulting in an escalation of high-risk assets and subpar clientele. A significant risk-loss event will disseminate risk throughout the financial system via the credit chain, intensifying systemic risk. Consequently, Hypothesis 4 is confirmed.

4.4. Heterogeneity Test

4.4.1. Heterogeneity of Economic Development Level

Countries exhibiting varying degrees of economic development demonstrate disparities in the risk and distribution frameworks of investment projects, while banks and other financial institutions possess markedly different tolerances for project failure risk, thereby influencing the impact of LOLR support on systemic financial risk. This paper examines the varying effects of LOLR support on financial systemic risk across countries with differing economic development levels, utilizing the classification established in the 2023 IMF World Economic Outlook to categorize the sample countries into Advanced Economies and Emerging Market and Developing Economies (EMDEs). The regression findings presented in columns (1)–(2) of Table 7 demonstrate that the effect of LOLR support on systemic risk impact in advanced countries is insignificant. Conversely, LOLR support in the EMDEs cohort elevates financial systemic risk at the 10% significance threshold. The rationale may be that the financial systems of industrialized nations are more advanced and possess a well-defined and comprehensive LOLR framework, coupled with stronger oversight of high-risk enterprises, which can successfully mitigate systemic financial risk. EMDEs may prioritize economic growth, exhibit greater tolerance for investment project failures, and possess fragile financial systems susceptible to liquidity vulnerabilities, hence increasing systemic risk.

Table 7. Heterogeneity analysis.

	(1) AEs		(2) EMDEs	(3) Bank-led		(4) Market-oriented		(5) Low CR		(6) High CR
LOLR	−7.9998		0.0241 *	0.0382 ***		−1.8162		0.0449 ***		−8.6622
	(−1.5481)		(1.8538)	(3.7397)		(−0.1603)		(4.7361)		(−0.8703)
R2	0.1353		0.2959	0.2555		0.1842		0.2662		0.1771
N	1479		1494	1763		1210		1439		1534
		(7) Low RQ			(8) High RQ		(9) Low DT		(10) High DT
LOLR		0.0263 **			−6.3593		0.0227 **		−7.7695
		(2.2393)			(−0.8434)		(2.3812)		(−1.2115)
control		YES			YES		YES		YES
National FE		YES			YES		YES		YES
Time FE		YES			YES		YES		YES
R2		0.3374			0.1188		0.3293		0.1276
N		1329			1644		1415		1558

Note: ***, **, and * indicate significance at the 1%, 5%, and 10% levels, respectively. The t-values are in parentheses.

4.4.2. Financial Structure Heterogeneity

The configuration of a nation’s financial system dictates the channels of financial risk transmission and accumulation, hence influencing the impact of LOLR support on systemic financial risk. This paper categorizes sample countries into two groups—market-dominated financial systems and bank-dominated financial systems—to examine the influence of LOLR support on systemic risk across varying financial structures (Langfield & Pagano, 2016; Lin et al., 2023). The estimation results in columns (3)–(4) of Table 7 indicate that the explanatory coefficient of LOLR support on systemic risk for bank-dominated nations is considerably positive at the 1% significance level, whereas the link for market-dominated countries is insignificant. Countries with bank-dominated financial systems exhibit a significant reliance on the bank credit network and anticipate being “too big to fail” and “too interconnected to fail.” Consequently, they opt to persist in taking risks even after receiving LOLR support, which heightens the potential for losses and accelerates diffusion primarily through interbank linkages, rendering the financial system more susceptible to risk accumulation. Conversely, a market-driven financial system with robust direct financing channels has enhanced risk diversification capabilities, thereby anticipating a reduction in the adverse effects of the central bank’s LOLR interventions.

4.4.3. Capital Adequacy Heterogeneity

The level of bank capital serves as a risk buffer and necessarily influences the impact of LOLR support on systemic risk. This paper examines the varying effects of LOLR support on systemic risk in nations with differing capital adequacy levels by categorizing the sample countries into two groups—high capital reserves and low capital reserves—based on the median capital adequacy level of each country’s banking system. The regression results presented in columns (5)–(6) of Table 7 demonstrate that LOLR support considerably exacerbates systemic risk in nations with low capital adequacy, but the link is minor for nations with high capital adequacy levels. Insufficiently capitalized banks may lack adequate risk resilience, rendering their capital reserves inadequate to absorb losses from asset expansion, hence increasing the likelihood of systemic risk (Fang & Liu, 2023). Conversely, nations with elevated capital adequacy ratios possess greater loss-absorbing capacity, and stringent capital regulations help ensure that financial institutions operate prudently.

4.4.4. Heterogeneity of the Regulatory Environment

The regulatory quality of a nation directly influences the efficacy of LOLR implementation. This paper employs the regulatory quality data from the World Bank’s global governance database as a proxy to assess a country’s regulatory environment. The sample countries are classified into high and low regulatory quality categories based on the mean regulatory quality, and the paper subsequently examines the variation in the LOLR’s impact on systemic financial risk across countries with differing regulatory quality. The regression outcomes are presented in columns (7)–(8) of Table 7, demonstrating that LOLR support considerably exacerbates systemic risk in countries with low regulatory quality, but the results for countries with good regulatory quality are insignificant. The regression results indicate that LOLR support in countries with low regulatory quality may foster excessive risk-taking among financial institutions due to insufficient regulatory constraints, thereby increasing systemic risk accumulation. Conversely, the ex-ante risk assessment and ex post accountability mechanisms in countries with high regulatory quality are more robust, which can mitigate the influence of LOLR support on systemic risk.

4.4.5. Heterogeneity in Digital Technology Levels

The advancement of digital technology may raise the efficiency of financial information transmission, and a high degree of digital technology can improve the accurate execution of LOLR policies through real-time monitoring, thereby altering the effect of LOLR support on systemic risk. This paper employs the Global Digital Economy Development Index Report (TIMG) to assess disparities in digital technology development among nations, utilizing the median TIMG index to categorize the sample countries into those with high and low levels of digital technology development. The index integrates the four dimensions of digital technology, digital market, digital infrastructure development, and digital governance across nations. The regression results presented in Table 7 columns (9)–(10) demonstrate that the influence of LOLR assistance on the intensification of financial systemic risk becomes insignificant with the advancement of digital technology. Digital technology fortifies the information disclosure of financial institutions and improves the market’s capacity for information collection. Nations with advanced digital technology development mitigate financial institutions’ risks by diminishing information asymmetry. Information asymmetry is specifically manifested in three dimensions: the first pertains to the relationship between banks. The advancement of digital technology mitigates interbank market failures, enabling a bank with robust fundamental information but a temporary liquidity shortfall to procure funds from other banks at a fair cost, hence alleviating liquidity challenges. The second pertains to the relationship between banks and markets. The market, augmented by digital technology, improves information monitoring capabilities, compelling banks to regulate their conduct, provide accurate and truthful data, and mitigate excessive risk-taking behavior. In an economy with underdeveloped digital technology, market oversight inadequately detects banks’ misinformation due to technological shortcomings, resulting in a partial breakdown of supervision and facilitating the banking sector’s exploitation of the LOLR bailout for arbitrage activities. The third relationship is between commercial banks and the central bank. The central bank can leverage digital technology to ascertain the genuine hardship of institutions seeking a bailout, differentiate between those experiencing liquidity shocks and insolvent entities, thereby facilitating targeted assistance. Simultaneously, the central bank may employ technical methods to monitor the allocation of LOLR rescue funds to mitigate financial system risk. Consequently, the advancement of digital technology significantly influences the efficacy of the LOLR strategy.

5. Conclusions

Traditional views hold that the LOLR’s provision of liquidity support to financial institutions during crises can prevent risk contagion, playing a crucial role in maintaining financial stability. However, the effectiveness of the LOLR in preserving financial stability during stable economic periods remains controversial. To address this, this paper analyzes the relationship and operational mechanisms between central bank LOLR support and systemic financial risks during stable economic periods, based on quarterly data from 55 countries worldwide. The findings reveal: First, a significant positive correlation exists between LOLR support and systemic financial risk. Specifically, during stable economic periods, the routine deployment of central bank LOLR support correlates with an escalation of systemic financial risk. This outcome is closely tied to moral hazard effects. Frequent or predictable liquidity support may weaken financial institutions’ risk discipline, incentivizing them to engage in riskier investment behaviors and thereby amplifying systemic vulnerability.

Second, this study confirms that LOLR support primarily amplifies systemic risk by influencing financial institutions’ asset allocation behavior, specifically through three channels: the asset liquidity erosion mechanism, the financial leverage intensification mechanism, and the asset quality deterioration mechanism. During stable economic periods, LOLR encourages increased allocation to illiquid long-term assets, exacerbating maturity mismatches and sowing systemic liquidity risks. This exacerbates maturity mismatches and sows systemic liquidity risks. Additionally, anticipating central bank liquidity bailouts, financial institutions may increase debt financing and expand leverage, amplifying risk exposure and contagion. Finally, aggressive asset allocation and high-leverage operations deteriorate asset quality. When the economy slows, latent non-performing assets become visible and spread through credit chains, eroding financial stability. Empirical tests in this study confirm these transmission channels, indicating that the policy effectiveness of LOLR depends not only on its design but also on financial institutions’ behavioral responses.

Third, the LOLR’s systemic risk amplification effect exhibits significant heterogeneity, particularly pronounced in emerging markets and developing economies, banking-dominated financial systems, and countries with low regulatory quality, limited digital technology development, and inadequate capital adequacy. The finding deepens our understanding of LOLR effectiveness across varying levels of economic development and institutional environments. This further underscores that financial stability policies must fully consider country-specific characteristics and institutional foundations, implementing differentiated and nuanced policy execution.

Based on this study’s core findings regarding the negative effects of LOLR and the significant cross-country heterogeneity, the following policy recommendations are proposed. First, policy design should directly target risk-generating channels. To curb financial institutions’ risk asset allocation after receiving liquidity support, regulators can link eligibility or costs for LOLR access to the liquidity structure and risk profile of their asset portfolios. For instance, imposing higher funding rates or stricter collateral requirements on banks with excessively high long-term asset ratios would constrain maturity mismatches and risk accumulation at the source.

Second, translate the scale or frequency of central bank liquidity support into additional countercyclical capital or systemic risk surcharges, thereby reducing financial institutions’ incentive to take risks. Making implicit subsidies explicit and costly guides institutions to internalize capital costs associated with risk-taking behavior in advance, proactively curbing excessive leverage impulses.

Third, building on this foundation, policies must further implement fine-tuned matching based on country-specific heterogeneity. For emerging and developing economies, the LOLR should be designed as a “strongly conditional” contract featuring rigorous ex-post verification and transparent fund flows to strengthen budget constraints and enhance transparency. Additionally, central banks should establish robust risk assessment and accountability mechanisms to accurately identify institutions facing temporary liquidity crunches versus those in substantive insolvency. For non-systemically important “zombie institutions” that chronically rely on bailouts, initiate market-based exit procedures to sever risk transmission chains. In countries with lagging digital technology and financial infrastructure, central banks should proactively build risk monitoring data platforms to accurately identify liquidity distress and solvency issues in financial institutions. This enables targeted rescue operations, thereby reducing the financial system’s expectation of LOLR bailouts.

This study has the following limitations that warrant further refinement in future research. The focus on “periods of economic stability” implies that the findings may not directly apply to financial crises or severe liquidity crunches. Under extreme stress scenarios, the stabilizing function of the LOLR as a “last line of defense” may exhibit complex interactions with the “moral hazard” effects revealed in this paper. Future research could extend the sample period to include major financial crises. Through comparative analysis, it could systematically examine the differential impact of the LOLR on financial stability under “crisis mode” versus “stable mode,” thereby providing more comprehensive decision-making guidance for central banks deploying this tool under varying macroeconomic conditions. Additionally, although this paper references mainstream literature in quantifying LOLR, the availability of indicators makes it difficult to precisely distinguish whether it serves as a bailout for temporary liquidity difficulties or as a liquidity supply potentially used to sustain daily operations. Future research can build upon this foundation for further refinement.