The Effect of Economic Policy Uncertainty on Banks: Distinguishing Short- and Long-Term Effects

Abstract

The interplay between government economic policy uncertainty (EPU) and bank risk remains a key concern in the financial stability literature. This study advances the field by examining the dynamic, time-varying impact of EPU on bank risk, explicitly differentiating between short- and long-term effects. We posit a dual hypothesis: heightened EPU increases short-run bank risk by raising borrower default probabilities while decreasing long-run risk as banks adopt more conservative lending strategies, given the option value of waiting under high uncertainty. Analyzing bank-level data across 22 countries from 1998 to 2017, we find robust empirical support: EPU exerts an immediate positive effect on bank risk and a significant negative effect with a lag of two to four years. These findings are robust to endogeneity and multiple sensitivity checks. Our results explicitly demonstrate the dual role of policy uncertainty in shaping bank risk-taking and offer timely guidance for the design of regulatory and macroprudential frameworks.

1. Introduction

Recent research increasingly recognizes that uncertainty in government economic policies has significant implications for businesses, the financial sector, and households (Schweitzer and Shane 2011; Giavazzi and McMahon 2012; Pastor and Veronesi 2013; Brogaard and Detzel 2015; Ghosal and Ye 2015; Gulen and Ion 2016; Julio and Yook 2016; Krol 2017; Xu 2023). Focusing on banking, prior studies document that heightened government policy uncertainty influences key bank decisions, resulting in higher loan interest rates (Francis et al. 2014; Ashraf and Shen 2019; Ashraf 2021), reduced lending activity (Bordo et al. 2016; Caglayan and Xu 2018; Hu and Gong 2019; Danisman et al. 2020; Shabir et al. 2022), increased liquidity hoarding (Ashraf 2020; Berger et al. 2022), higher bank risks (Nguyen 2021; Phan et al. 2021; Shabir et al. 2021; Ali et al. 2023; De Silva et al. 2023; Danisman and Tarazi 2024; Moudud-Ul-Huq and Akter 2024; Samarasinghe and Ahmed 2025) and lower earnings quality (Yiqiang Jin et al. 2019; Ng et al. 2020; Danisman et al. 2021; Desalegn and Zhu 2021; Tran and Houston 2021; Biswas et al. 2025), ultimately leading to lower bank returns (Ashraf 2025).

These studies rest on the premise that heightened government economic policy uncertainty (EPU) deteriorates the macroeconomic environment, elevates risk and default probabilities of economic agents, and induces greater risk aversion. While EPU exerts immediate effects, it also generates lagged responses as the option value of waiting rises under uncertainty. Stokey (2016) demonstrates that uncertainty surrounding policy changes prompts firms to adopt a “wait-and-see” strategy, delaying irreversible investments until uncertainty resolves. Once clarity emerges, firms often accelerate investment, illustrating the dynamic and delayed impact of uncertainty on decision-making. In the banking context, institutions may initially curtail lending following a surge in uncertainty yet adopt more aggressive strategies over time, ultimately increasing credit exposure. Likewise, conservative lending during periods of elevated uncertainty may reduce credit risk in subsequent periods. In this paper, we examine the impact of EPU on bank risk, explicitly distinguishing between short- and long-term effects.

We hypothesize that economic policy uncertainty can affect bank risk in two ways: First, heightened policy uncertainty is likely to boost current-period bank risk by raising the average default risk of borrowers. The average default risk of borrowers increases because policy uncertainty shocks lead to higher idiosyncratic dispersion in firms’ productivity (Brand et al. 2019) and household incomes (Bloom 2014; Li et al. 2018; Lee et al. 2021; Gu et al. 2022; Zhang and Ling 2024) due to a decrease in overall economic activity, including new investment, employment, and household consumption (Bloom 2009; Baker et al. 2016; Bloom et al. 2018). Higher idiosyncratic dispersion in incomes enhances the probability of a bad state for both borrowing firms and households.

Second, the real-options theory-based literature suggests that banks reduce lending in response to higher uncertainty, which increases the value of the option to wait and see (Bordo et al. 2016; Caglayan and Xu 2018; Hu and Gong 2019; Danisman et al. 2020; Shabir et al. 2022), while some other studies suggest a reduction in current-period bank lending might result in lower bank default risk and loan losses with a lag of two to four years (Hess et al. 2009; Foos et al. 2010; Amador et al. 2013). Building on these studies, our second hypothesis is that current-period economic policy uncertainty, because of its negative impact on current-period bank lending, might result in lower bank risk with a lag of two to four years.

To investigate the impact of economic policy uncertainty on bank risk with respect to both hypotheses, we use bank-level data from 22 countries over the period from 1998 to 2017 for empirical analysis. We measure bank risk with five alternative proxies, including the probability of bank default (z-score), volatility in bank total operating income, volatility in bank net interest margins, non-performing loans to gross loans ratio, and loan loss provisions to gross loans ratio. These measures capture different aspects of bank risk. We measure government economic policy uncertainty with two alternative proxies, including the world uncertainty index of Ahir et al. (2018) and the economic policy uncertainty index of Baker et al. (2016). Overall, our results provide evidence in favor of both hypotheses. The results remain robust when we use the political fractionalization index as an instrument for policy uncertainty, alternative estimation methods, alternative sample compositions, and alternative control variables.

We contribute to the existing literature in several ways: First, we add to the recently expanding literature that argues for the importance of economic policy uncertainty for economic outcomes. In this regard, existing studies have found that during the periods of higher policy uncertainty in a country, firms invest less (Bernanke 1983; Bloom 2009; Baker et al. 2016), firms delay merger and acquisition deals (Bonaime et al. 2018), foreign direct investment shrinks (Julio and Yook 2016), unemployment swells (Baker et al. 2016), and the gross domestic product drops significantly (Bloom et al. 2018).

Second, we complement the studies that argue that economic policy uncertainty is important for financial sector outcomes. In this regard, recent studies have found that firms pay higher costs on corporate bonds (Waisman et al. 2015; Bradley et al. 2016) and equity capital (Pastor and Veronesi 2013; Brogaard and Detzel 2015; Pham 2019) as economic policy uncertainty increases. Specifically focusing on the banking sector, the recent literature reports that bank loan interest rates increase (Francis et al. 2014; Ashraf and Shen 2019), loan growth declines (Bordo et al. 2016; Caglayan and Xu 2018; Hu and Gong 2019; Danisman et al. 2020; Shabir et al. 2022), liquidity hoarding increases (Ashraf 2020; Berger et al. 2022), and stock prices plummet (He and Niu 2018) in response to higher economic policy uncertainty. We complement these studies by investigating the impact of economic policy uncertainty on bank risk for a sample of 22 countries. Our main result, in this regard, is that economic policy uncertainty leads to higher current-period bank risk, while a lower bank risk with a lag of 2–4 years.

Finally, we add to the studies that examine the country-level environmental determinants of bank risk. In this regard, existing studies find that banking industry regulations such as minimum capital requirements, restrictions on bank activities and the existence of explicit deposit insurance (Laeven and Levine 2009; Anginer et al. 2014; Haq et al. 2014; Ashraf et al. 2020), political institutions (Ashraf 2017; Dutra et al. 2024), legal institutions such as creditor rights and borrowers’ related information-sharing mechanisms (Houston et al. 2010; Cole and Turk 2013; Fang et al. 2014), national culture (Ashraf et al. 2016a; Ashraf and Arshad 2017; Mourouzidou-Damtsa et al. 2019; Illiashenko and Laidroo 2020), and trade and financial openness (Ashraf et al. 2017; Ashraf 2018; Rahman et al. 2020) are significant determinants of bank risk. We find that economic policy uncertainty is also an economically significant determinant of bank risk.

The paper is organized as follows. In Section 2, we draw testable hypotheses. Section 3 presents our data collection procedures. Section 4 introduces empirical methodology and variables. Empirical results are reported in Section 5. Section 6 concludes the study.

2. Literature Review and Hypothesis Development

We postulate that economic policy uncertainty boosts bank risk immediately by raising the default risk of borrowers, while it reduces bank risk with a lag of two to four years by decreasing current-year loan growth.

The banking sector is one of the main sources of financing for the real sector (i.e., businesses and households) in all major economies. Banks provide loans to earn interest; however, they face the risk of whether borrowers will be able to pay back the principal and interest payments. Thus, bank risk, in large part, depends on the factors that affect the financial conditions of borrowers. Keeping other things constant, the deterioration in the financial conditions of borrowers would adversely affect their debt-paying ability and, consequently, would increase the bank risk.

Economic policy uncertainty negatively impacts the real sector. When uncertainty rises, risk-averse economic agents tend to reduce investment and consumption. The recent literature shows that elevated economic policy uncertainty dampens short-term economic growth, both quarterly and annual, by lowering investment, hiring, output, consumption, and trade (Bloom 2014, 2017). While economic agents become more cautious in making new decisions, heightened uncertainty also increases default risk on existing commitments, such as outstanding loans, by expanding the left tail of default outcomes for both businesses and households. For firms, greater policy uncertainty raises default risk by increasing idiosyncratic dispersion in productivity (Bloom 2009; Liu and Zhong 2017; Brand et al. 2019). For households, it amplifies income volatility through reduced hiring and higher unemployment on one hand and wage fluctuations among those still employed on the other (Bloom 2014; Li et al. 2018; Lee et al. 2021; Gu et al. 2022; Zhang and Ling 2024). A sustained rise in aggregate economic policy uncertainty gradually deteriorates overall economic conditions, ultimately causing financially vulnerable businesses and households to default on bank loans. Consequently, economic policy uncertainty increases bank risk by worsening ex post loan performance. Recent empirical evidence largely supports a positive association between policy uncertainty and bank risk (Nguyen 2021; Phan et al. 2021; Shabir et al. 2021; Ali et al. 2023; De Silva et al. 2023; Danisman and Tarazi 2024; Moudud-Ul-Huq and Akter 2024; Samarasinghe and Ahmed 2025). Based on this reasoning, our first hypothesis is as follows:

Hypothesis 1.

Economic policy uncertainty boosts bank risk immediately.

Economic policy uncertainty can indirectly affect bank risk by reducing loan growth. Building on the real-options theory, studies such as Bernanke (1983), Bloom (2009), and Baker et al. (2016) find that heightened uncertainty reduces new investments. In a similar vein, higher economic policy uncertainty can reduce bank investment in new loans by increasing the value of the option to wait because of the chances of unfavorable policies. Unfavorable policies, such as adverse regulations or unexpected changes in government expenditures or interest rates, might lead to weaker economic conditions. Thus, in the face of heightened economic policy uncertainty, banks would hesitate to lend and postpone real investments to retain the option value. Consistent with these arguments, Bordo et al. (2016), Chi and Li (2017), Hu and Gong (2019), Danisman et al. (2020), and Shabir et al. (2022) find that economic policy uncertainty and bank loan growth rates have a generally negative association.

Although higher EPU arguably reduces lending through banks’ precautionary “wait-and-see” behavior, lower lending may also result from banks adopting more conservative credit standards. The prior literature emphasizes a strong relationship between lending growth and credit standards. Banks can stimulate lending by relaxing collateral requirements, lowering interest rates, or employing a combination of both (Dell’Ariccia and Marquez 2006). For example, Dell’Ariccia et al. (2012) find that, prior to the subprime financial crisis of 2007–2009, loan denial rates were relatively low, and lenders placed less emphasis on applicants’ loan-to-income ratios in U.S. regions experiencing higher lending growth. Recent evidence suggests that banks tighten lending standards during periods of higher policy uncertainty. For instance, Bordo et al. (2016) show that U.S. banks adopted stricter lending standards under heightened uncertainty. Similarly, Ashraf and Shen (2019) report that banks tend to charge higher interest rates on loans during such periods.

The decline in loan growth, driven by both precautionary “wait-and-see” behavior and stricter credit standards, may have significant implications for bank risk in subsequent periods. For instance, in a seminal paper, Foos et al. (2010) argue that loan growth is an important driver of bank risk with lagged effects. Using bank-level data from 16 countries, they demonstrate that excessive current-period loan growth is associated with an increase in loan loss provisions over the subsequent three years. Similarly, Hess et al. (2009) use data from 32 Australian banks over the period 1980–2005 and find that strong loan growth is associated with higher loan losses, with a lag of two to four years. Likewise, Amador et al. (2013) use the data of Colombian banks and find that abnormal credit growth over a prolonged period results in higher bank default risk and non-performing loans in subsequent periods. In another recent study, Dang (2019) used the data of Vietnamese banks over the period 2006–2017 and concluded that higher loan growth leads to an increase in loan loss provisions during the subsequent 2 to 3 years.

Building on the above discussion, our second hypothesis is that economic policy uncertainty reduces bank risks, such as default risk, loan loss provisions, and non-performing loans, with a lag of two to four years due to its immediate negative impact on loan growth.

Hypothesis 2.

Economic policy uncertainty reduces bank risk with a lag of two to four years.

3. Data Collection

We started our sample construction by downloading the data of the world uncertainty index of Ahir et al. (2018) and the economic policy uncertainty index of Baker et al. (2016) from the website http://www.policyuncertainty.com. We chose the countries for which both of these indices are available. We chose a fairly long sample period from 1998 to 2017. Next, we obtained financial statement accounting data for banks in these countries from the BankFocus database. We included commercial, cooperative, and savings banks in the sample. We kept the data of both active and inactive banks to avoid survivorship bias of low-risk banks.

Next, we collected data on banking industry-level and country-level control variables. We collected data on banking industry regulations from World Bank surveys on bank regulations (Barth et al. 2013) and data on the banking industry structure from the Financial Development database of the World Bank. Data for macroeconomic variables was collected from the World Development Indicators (WDI) database of the World Bank. Data for the law and order variable was collected from the International Country Risk Guide (ICRG) database. Then, we linked bank-level financial statements data with the banking industry-level and country-level data.

To refine the data, we dropped observations with missing necessary data. We also dropped banks with fewer than four valid yearly observations. We excluded U.S. banks from the sample to ensure that their high volume of observations did not bias the cross-country results. Our final dataset consists of 5138 banks with 50,595 yearly observations from 22 countries over the period 1998–2017. We winsorized all bank-level variables at the 1% level in both the lower and upper tails to eliminate the effect of outliers.

Table 1. Sample distribution and country-wise mean values of main variables.

			Country-Level Mean Values
Country	Banks	Observations	Z-Score	σ(ROA)	σ(NIM)	LLP	NPL	WUI	EPU
Australia	31	200	−4.39	0.17	0.15	0.34	1.12	0.19	4.74
Brazil	115	1016	−2.75	1.60	2.03	2.88	8.01	0.27	4.96
Canada	58	287	−4.14	0.24	0.22	0.42	1.54	0.15	5.02
Chile	24	154	−3.80	0.45	0.53	1.19	3.72	0.15	4.70
China	187	1185	−3.73	0.30	0.42	0.97	1.74	0.10	5.23
Colombia	30	209	−3.24	0.94	0.88	2.60	3.18	0.26	4.61
France	270	2399	−4.07	0.27	0.23	0.49	5.28	0.20	5.08
Germany	1888	23,442	−3.92	0.21	0.18	0.54	3.00	0.17	4.84
Greece	18	125	−3.07	0.57	0.34	2.23	18.49	0.12	4.71
Hong Kong	33	295	−4.23	0.28	0.22	0.36	0.61	0.10	4.92
India	72	643	−3.75	0.32	0.29	1.20	4.24	0.10	4.52
Ireland	9	69	−3.59	0.29	0.20	1.69	15.00	0.23	4.83
Italy	661	4906	−3.76	0.35	0.29	1.16	12.17	0.22	4.75
Japan	638	8797	−4.20	0.13	0.09	0.49	7.36	0.18	4.64
Mexico	158	892	−3.19	0.96	1.06	2.48	2.83	0.26	4.06
Netherlands	37	253	−3.59	0.37	0.28	0.82	3.88	0.22	4.56
Rep. Korea	17	85	−3.65	0.44	0.34	0.85	1.83	0.19	4.93
Russia	515	2775	−2.96	1.41	1.37	2.13	7.80	0.25	5.05
Singapore	11	86	−4.75	0.14	0.23	0.21	1.22	0.06	4.87
Spain	163	981	−3.62	0.36	0.24	0.78	6.84	0.26	4.73
Sweden	93	910	−4.25	0.33	0.29	0.28	1.91	0.20	4.50
United Kingdom	110	886	−3.74	0.41	0.36	0.90	5.29	0.37	4.82
Total/mean	5138	50,595	−3.86	0.34	0.31	0.80	6.42	0.19	4.80

This table reports the country-wise sample distribution and mean values of the dependent and main independent variables. Z-score, σ(ROA), σ(NIM), LLP, and NPL are five alternative proxies of bank risk, where higher values for each of these proxies represent higher bank risk and vice versa. WUI and EPU are two alternative proxies of government economic policy uncertainty. WUI is the world uncertainty index of Ahir et al. (2018). EPU is the news-based economic policy uncertainty index of Baker et al. (2016). Higher values for both WUI and EPU represent higher policy uncertainty and vice versa.

reports the countries included in the sample, together with the number of banks and annual observations from each sample country.

4. Empirical Methodology

Following recent cross-country studies on banking (Ashraf 2017; Ashraf and Shen 2019), we specify the following pooled-panel OLS model for estimation. The pooled-panel model takes into account both cross-country and time variations and is considered superior for cross-country studies involving country-level variables (Ashraf 2017).

$$Y_{i,j,t}={\alpha }_i+{\beta }_1{Policy Uncertainty}_{j,t}+\sum _{k=1}^k{\beta }_k{X}_{i,j,t}^k+\sum _{l=1}^l{\beta }_l{X}_{j,t}^l+\sum _{m=1}^m{\beta }_m{X}_{j,t}^m+\sum _{t=1}^{T-1}{ϵ}_t{D}_t+{\epsilon }_{i,j,t}$$

(1)

Here, i, j, and t subscripts represent bank, country, and year, respectively. α_i is a constant term. The dependent variable, Y, represents bank risk. Policy uncertainty represents the government’s economic policy uncertainty and is the main explanatory variable of interest. $X_{i,j,t}^k$ represents bank-level annual control variables, including bank size, the equity-to-total assets ratio, the non-interest income-to-total revenue ratio, annual growth in bank gross loans, the liquid assets-to-total assets ratio, the cost-to-income ratio, the list bank dummy, the state-owned bank dummy, and the bank market power ratio. $X_{j,t}^l$ represents banking-industry-level regulatory and industry-structure control variables, including capital stringency index, activity restrictions, and banking industry concentration. $X_{j,t}^m$ represents country-level macroeconomic and institutional control variables, including GDP growth rate, inflation, law and order, and the financial crisis dummy. D_t is a set of year dummy variables to control for international factors. ε_i,j,t is an error term. We use heteroskedastic-robust standard errors to estimate p-values in regressions.

To examine the first hypothesis, we estimate Equation (1), and the estimates of ${\beta }_1$ show the immediate impact of economic policy uncertainty on bank risk. For the second hypothesis, we estimate Equation (1) by using lags of policy uncertainty together with all other explanatory variables. We use up to four lags in alternative estimations. Coefficient estimates of lagged policy uncertainty show that the impact of policy uncertainty on bank risk occurs with a lag.

We measure bank risk with five alternative proxies, including Z-score, σ(NIM), σ(ROA), LLP, and NPL.

The Z-score measures the probability of bank default. Z-score = −1 × (log((ROA + CAR)/σ(ROA))), where ROA is the annual return on assets before loan loss provisions and taxes, CAR is the annual equity-to-total assets ratio, and σ(ROA) is the standard deviation of annual values of return on assets before loan loss provisions and taxes calculated over a 3-year overlapping window over the sample period (i.e., 1998–2000, 1999–2001, and so on). The values of the Z-score show the number of standard deviations from the mean value by which the bank return has to fall to deplete all shareholders’ equity. The higher the values of the Z-score, the higher the banks’ default risk, and vice versa. Lepetit and Strobel (2015) showed that the Z-score defines bank default risk over the domain of all real numbers and is an attractive and unproblematic proxy of bank default risk to be used as a dependent variable in standard regression analysis.

σ(NIM) measures the volatility in bank net interest margins. Specifically, σ(NIM) equals the standard deviation of annual net interest margin, calculated over 3-year overlapping periods (i.e., 1998–2000, 1999–2001, and so on). σ(NIM) denotes bank interest income risk.

σ(ROA) measures the volatility in overall bank operating income. Specifically, σ(ROA) equals the standard deviation of annual values of return on assets before loan loss provisions and taxes, calculated over 3-year overlapping periods. σ(ROA) represents the bank’s overall operating income risk. Due to the 3-year overlapping window used for the calculation of σ(NIM) and σ(ROA), the effective sample period for empirical analysis starts from the year 2000.

The LLP variable equals the annual loan loss provisions-to-gross loans ratio. As banks regularly adjust loan loss provisions to reflect the changes in the risk of outstanding loans (Laeven and Huizinga 2019), an increase in loan loss provisions represents higher bank loan portfolio risk.

NPL equals the annual non-performing loans-to-gross loan ratio for each bank. NPL represents bank-realized credit losses. According to the United Nations System of National Accounts, a loan is classified as non-performing when payments of interest or principal are past due by 90 days or more. Banks periodically update their stock of non-performing loans if they find that certain borrowers have not paid back on their bank loans. Thus, the higher the volume of non-performing loans, the higher the bank’s realized risks.

The main advantage of using alternative proxies is that they capture different aspects of bank risk. The Z-score measures bank default risk, σ(NIM) and σ(ROA) capture volatility in bank net interest income and total earnings, and LLP and NPL represent bank expected and realized credit losses.

We use two alternative proxies to measure government economic policy uncertainty: the world uncertainty index of Ahir et al. (2018) and the economic policy uncertainty index of Baker et al. (2016). Ahir et al. (2018) developed the world uncertainty index (WUI hereafter) by counting the word “uncertainty” (or its variants) in the quarterly country analysis reports prepared by the Economist Intelligence Unit (EIU) of the Economist Group for all countries around the world. These reports are prepared by a group of experts and discuss country-specific political and economic developments. We averaged quarterly values of the WUI index for each year to obtain annual values to be used with annual bank-level data. Different from Ahir et al. (2018), Baker et al. (2016) construct the monthly economic policy uncertainty index (EPU hereafter) based on the count of newspaper articles containing keywords in three categories, including uncertainty (i.e., uncertain or uncertainty), economy (i.e., economic or economy), and policy (i.e., central bank, regulation, tax, government spending, or other country-specific policy-related words), published in major newspapers of each country. We averaged the monthly EPU index for annual values.

Ahir et al. (2018) argue that the WUI index is based on country reports from a single source and better captures local political and economic policy uncertainty as compared to the EPU index, which is more global in nature. Newspaper articles counted for the EPU index may also include those articles that discuss uncertainty related to international factors. Because of this, the EPU index is more likely to co-move internationally; international factors explain 36 percent of the variation in the EPU index but only 17 percent in the WUI index (Ahir et al. 2018). Further, the WUI index does not suffer from concerns such as ideological bias and consistency and makes cross-country comparisons easier due to the underlying single data source.

We add several variables to control for bank-level, banking industry-level, and country-level factors, which are likely to affect bank risk in addition to economic policy uncertainty.

Following the previous literature (Ashraf et al. 2016a; Ashraf 2017; Zheng et al. 2017; Moudud-Ul-Huq et al. 2018; Nguyen 2021; Phan et al. 2021; Shabir et al. 2021; Ali et al. 2023; De Silva et al. 2023; Danisman and Tarazi 2024; Moudud-Ul-Huq and Akter 2024; Samarasinghe and Ahmed 2025), we add bank-level variables to control for individual bank characteristics, such as size, capitalization, growth, business model, liquidity holding, efficiency, ownership structure, and market power, which are likely to have a significant effect on bank risk. Specifically, we include bank size, growth in bank gross loans, the non-interest income-to-total revenue ratio, the liquid assets-to-total assets ratio, the cost-to-income ratio, the listed banks dummy, the government banks dummy, and bank market power. Bank size is measured as the natural logarithm of bank annual total assets. Growth in bank gross loans is measured as the year-on-year growth in bank total gross loans. The non-interest income-to-total revenue ratio measures non-interest income, including net gains on trading and derivatives, net gains on other securities, net fees and commissions, and other operating income, as a percentage of total bank revenue. The liquid assets-to-total assets ratio measures the year-end balances of liquid assets (cash and due from banks, trading securities and at fair value through income, loans and advances to banks, reverse repos, and cash collaterals) as a percentage of total bank assets. The cost-to-income ratio measures the cost of running operations as a percentage of a bank’s operating income. The listed banks dummy equals 1 if a bank is listed on a stock exchange and 0 otherwise. The government bank dummy equals 1 if the government holds the majority shareholding of a bank and 0 otherwise. Bank market power is measured as the annual total assets of an individual bank divided by the sum of assets of all banks operating in a country in a year, multiplied by 100.

We add three banking industry-level variables to control for the effect of prudential regulations and banking industry structure on bank risk: capital stringency index, activity restrictions, and banking industry concentration. Data for the capital stringency index and activity restrictions is obtained from World Bank surveys on bank regulations as reported by Barth et al. (2013). Capital stringency index equals the sum of two sub-indices, including the initial capital stringency index and the overall capital stringency index. Overall, the capital stringency index reflects whether the minimum capital requirements for banks in a country are in line with Basel requirements; whether the minimum capital requirements are sensitive to bank credit, market, and operational risks; and whether regulators verify the sources of bank capital. Further, it also measures which types of funds can be categorized as bank capital and which types of losses banks have to deduct to determine capital adequacy ratios. The index ranges from 0 to 10, where higher values indicate that a country implements stringent capital requirements for banks, and vice versa. The activity restrictions variable measures which commercial banks in a country are allowed to participate in non-lending activities such as securities, insurance, real estate activities, and/or owning other firms. The variable ranges from 4 to 16, where higher values indicate that a country implements higher restrictions on bank activities, and vice versa. As the World Bank surveys on bank regulations were conducted in 1999, 2003, 2007, and 2011, we use information from the survey conducted in 1999 for bank observations over the year 2000, from 2003 survey for bank observations over the years 2001–2003, from 2007 survey for bank observations over the years 2004–2007, and from 2011 survey for banks observations over the years 2008–2018. Banking industry concentration is measured as “the assets of the three largest banks as a percentage of the total assets of all banks operating in a country”. We collected data for this variable from the Financial Development database of the World Bank.

Finally, we add country-level macroeconomic and legal institutional variables to control for the impact of cross-country and over-time variations in macroeconomic conditions and legal institutions on bank risk. Macroeconomic variables include GDP growth rate, inflation, and the developing countries dummy. The GDP growth rate equals the annual percentage growth in gross domestic product of a country. Inflation equals the percentage change in the annual average consumer prices. The developing countries dummy equals 1 if a sample country is classified as middle or low income by the World Bank and 0 if it is classified as developed. Data for these variables was collected from the World Development Indicators database of the World Bank. We include the law and order variable, which measures the extent of law enforcement in a country, to account for cross-country differences in the legal institutional environment. Further, a financial crisis may increase bank risk by materializing downside risks. To control for this effect, we add a financial crisis dummy variable, which equals 1 if a country faces a financial crisis in a year according to Laeven and Valencia’s (2018) financial crisis database, and 0 otherwise.

5. Empirical Results

5.1. Summary Statistics

Table 1 reports the country-wise mean values of each of the five proxies of bank risk and two proxies of policy uncertainty. As shown, the mean Z-score is higher in emerging market countries such as Brazil (−2.75) and Russia (−2.96), indicating that banks, on average, have a higher probability of default in these countries. On the other hand, the mean Z-score is lower for countries, such as Singapore (−4.75) and Australia (−4.39), with stable financial systems. Likewise, the mean value of the WUI index is the highest for the United Kingdom (0.37), due to Brexit leading to uncertainty shock, and the lowest for Singapore (0.06).

Table 2. Summary statistics of main variables.

Variable	Observations	Mean	Standard Deviation	Minimum Value	Maximum Value
Z-score	50,595	−3.86	1.00	−8.15	3.79
σ(ROA)	50,595	0.34	0.58	0.01	9.79
σ(NIM)	50,595	0.31	0.62	0.00	12.18
LLP	50,595	0.80	1.41	−2.80	8.31
NPL	30,266	6.42	6.61	0.02	35.41
WUI	50,595	0.19	0.12	0.00	0.99
EPU	50,595	4.80	0.37	3.30	5.90
Bank size	50,595	14.06	2.02	6.04	22.11
Equity-to-total assets ratio	50,595	8.77	6.85	1.95	92.98
Non-interest income-to-total revenue ratio	50,595	26.46	17.15	−15.58	88.25
Annual growth in bank gross loans	50,595	5.89	16.34	−33.27	95.15
Liquid assets-to-total assets ratio	50,595	17.80	13.96	1.77	73.14
Cost-to-income ratio	50,595	69.51	15.95	27.92	129.12
Listed bank dummy	50,595	0.08	0.27	0.00	1.00
State-owned bank dummy	50,595	0.01	0.10	0.00	1.00
Bank market power	50,595	0.51	2.96	0.00	100.00
Banking industry concentration	50,595	61.51	15.17	21.84	100.00
Capital stringency index	50,595	6.63	1.48	2.00	10.00
Activity restrictions	50,595	8.34	2.12	4.00	15.00
Developing countries dummy	50,595	0.14	0.34	0.00	1.00
GDP growth rate	50,595	1.44	2.77	−7.80	25.01
Inflation	50,595	1.95	2.55	−3.74	21.46
Law and order	50,595	4.69	0.81	1.00	6.00
Financial crisis dummy	50,595	0.13	0.33	0.00	1.00

This table reports the summary statistics of the main variables. Z-score, σ(ROA), σ(NIM), LLP, and NPL are five alternative proxies of bank risk, where higher values for each of these proxies represent higher bank risk and vice versa. WUI and EPU are two alternative proxies of government economic policy uncertainty. WUI is the world uncertainty index of Ahir et al. (2018). EPU is the news-based economic policy uncertainty index of Baker et al. (2016). Higher values for both WUI and EPU represent higher policy uncertainty and vice versa. The others are bank-, banking industry-, and country-level control variables.

reports full-sample summary statistics for the main variables. The mean value of the Z-score equals −3.86 with a standard deviation of 1.0. These summary statistics are largely comparable with previous studies on bank risk, such as Kanagaretnam et al. (2014) and Ashraf (2017), who report mean Z-score values equal to −3.48 and −3.64, respectively. The mean values of the WUI and EPU indices are 0.19 and 4.10 with standard deviations of 0.12 and 0.37, respectively, pointing to considerable within-sample variation in policy uncertainty. Likewise, the control variables also demonstrate substantial variation across mean values.

Table 3. Matrix of pair-wise correlations between variables.

Panel 1: Pearson correlations between alternative proxies of bank risk
	Variables	(1)	(2)	(3)	(4)	(5)
(1)	Z-score	1.00
(2)	σ(ROA)	0.58	1.00
(3)	σ(NIM)	0.31	0.60	1.00
(4)	LLP	0.25	0.31	0.29	1.00
(5)	NPL	0.22	0.17	0.11	0.38	1.00
Panel 2: Pearson correlations between alternative measures of policy uncertainty
		(1)	(2)
(1)	WUI	1.00
(2)	EPU	0.31	1.00
Panel 3: Pearson correlations between control variables
		(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)	(13)	(14)	(15)	(16)	(17)
(1)	Bank size	1.00
(2)	Equity-to-total assets ratio	−0.32	1.00
(3)	Non-interest income-to-total revenue ratio	0.05	0.18	1.00
(4)	Annual growth in bank gross loans	0.04	0.04	0.08	1.00
(5)	Liquid assets-to-total assets ratio	−0.05	0.21	0.09	0.03	1.00
(6)	Cost-to-income ratio	−0.25	−0.06	0.02	−0.13	0.08	1.00
(7)	Listed bank dummy	0.43	−0.03	0.05	0.07	−0.06	−0.15	1.00
(8)	State-owned bank dummy	0.11	0.05	0.03	−0.00	0.06	−0.06	0.13	1.00
(9)	Bank market power	0.38	−0.03	0.08	0.04	0.03	−0.13	0.30	0.09	1.00
(10)	Banking industry concentration	−0.16	−0.12	0.18	−0.08	−0.29	−0.00	−0.24	−0.08	0.01	1.00
(11)	Capital stringency index	−0.01	0.02	0.23	0.04	−0.14	−0.11	−0.04	0.03	−0.02	0.30	1.00
(12)	Activity restrictions	0.27	−0.06	−0.27	0.02	0.03	−0.13	0.22	−0.00	0.05	−0.57	−0.21	1.00
(13)	Developing countries dummy	−0.01	0.39	0.12	0.24	0.28	−0.17	0.21	0.15	0.12	−0.47	0.07	0.18	1.00
(14)	GDP growth rate	0.12	0.00	−0.00	0.19	0.02	−0.18	0.11	0.06	0.08	−0.07	0.15	0.13	0.27	1.00
(15)	Inflation	−0.15	0.37	0.24	0.21	0.24	−0.09	0.11	0.10	0.07	−0.23	0.13	−0.16	0.72	0.09	1.00
(16)	Law and order	0.09	−0.37	−0.09	−0.16	−0.20	0.10	−0.12	−0.09	−0.05	0.38	0.10	−0.17	−0.78	−0.03	−0.55	1.00
(17)	Financial crisis dummy	−0.02	−0.01	0.03	−0.03	0.00	0.02	−0.03	−0.04	−0.01	0.06	0.12	−0.08	−0.09	−0.49	0.04	0.06	1.00

This table reports pair-wise Pearson correlations between variables. All correlations are significant at the 5 percent level except those in boldface. Z-score, σ(ROA), σ(NIM), LLP, and NPL are five alternative proxies of bank risk, where higher values for each of these proxies represent higher bank risk and vice versa. WUI and EPU are two alternative proxies of government economic policy uncertainty. WUI is the world uncertainty index of Ahir et al. (2018). EPU is the news-based economic policy uncertainty index of Baker et al. (2016). Higher values for both WUI and EPU represent higher policy uncertainty and vice versa. The others are bank-, banking industry-, and country-level control variables.

reports pair-wise correlations between variables. As shown in Panel 1, correlation coefficients between alternative proxies of bank risk are positive but not very high, which indicates that each proxy, to some extent, measures bank risk from a different perspective. Likewise, the 0.31 correlation coefficient between WUI and EPU in Panel 2 suggests the two proxies are different in measuring uncertainty about government policy. Correlations between control variables are also not very high, suggesting that the chances of multicollinearity in multivariate models are remote.

5.2. Policy Uncertainty and Bank Risk: Main Specifications

To analyze Hypothesis 1, we estimate Equation (1) using five alternative proxies of bank risk as the dependent variable one by one, first representing policy uncertainty with WUI and then with the EPU index.

As shown in

Table 4. Impact of policy uncertainty on bank risk: main specifications.

Variables	Z-Score	σ(ROA)	σ(NIM)	LLP	NPL
	Model (1)	Model (2)	Model (3)	Model (4)	Model (5)
WUI	0.510 ***	0.158 ***	0.246 ***	0.592 ***	6.768 ***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Bank-level control variables
Bank size	−0.037 ***	−0.018 ***	−0.010 ***	−0.039 ***	−0.504 ***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Equity-to-total assets ratio	−0.030 ***	0.020 ***	0.017 ***	−0.014 ***	0.035 ***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Non-interest income-to-total revenue ratio	0.006 ***	0.003 ***	0.000	0.002 ***	0.024 ***
	(0.000)	(0.000)	(0.791)	(0.000)	(0.000)
Annual growth in bank gross loans	0.002 ***	0.001 ***	0.002 ***	−0.002 ***	−0.065 ***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Liquid assets-to-total assets ratio	0.004 ***	0.002 ***	0.002 ***	0.003 ***	0.019 ***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Cost-to-income ratio	0.013 ***	0.004 ***	0.002 ***	−0.014 ***	0.003
	(0.000)	(0.000)	(0.000)	(0.000)	(0.119)
Listed bank dummy	−0.072 ***	0.000	−0.015	0.104 ***	−0.445 ***
	(0.000)	(0.995)	(0.130)	(0.000)	(0.000)
State-owned bank dummy	0.173 ***	−0.074 ***	−0.053 **	0.438 ***	4.096 ***
	(0.000)	(0.000)	(0.014)	(0.000)	(0.000)
Bank market power	0.003 **	0.002 **	−0.001	0.007 ***	0.140 ***
	(0.027)	(0.035)	(0.332)	(0.001)	(0.000)
Banking industry-level control variables
Banking industry concentration	0.004 ***	−0.001 ***	0.000	0.005 ***	−0.053 ***
	(0.000)	(0.000)	(0.105)	(0.000)	(0.000)
Capital stringency index	−0.013 ***	−0.023 ***	−0.021 ***	−0.072 ***	−0.090 ***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.001)
Activity restrictions	−0.018 ***	−0.029 ***	−0.033 ***	−0.020 ***	0.557 ***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Country-level control variables
Developing countries dummy	0.776 ***	0.455 ***	0.504 ***	1.037 ***	−5.538 ***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
GDP growth rate	−0.004	−0.011 ***	−0.010 ***	−0.110 ***	−0.280 ***
	(0.162)	(0.000)	(0.000)	(0.000)	(0.000)
Inflation	0.015 ***	0.012 ***	0.018 ***	−0.001	−0.092 ***
	(0.000)	(0.000)	(0.000)	(0.720)	(0.000)
Law and order	−0.198 ***	−0.095 ***	−0.144 ***	−0.399 ***	−2.935 ***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Financial crisis dummy	0.089 ***	−0.032 ***	−0.030 ***	0.030	1.215 ***
	(0.000)	(0.001)	(0.005)	(0.278)	(0.000)
Year FE	Yes	Yes	Yes	Yes	Yes
Constant	−3.083 ***	0.930 ***	1.177 ***	5.139 ***	25.090 ***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Observations	50,595	50,595	50,595	50,595	30,266
R-squared	0.251	0.403	0.390	0.228	0.255

This table reports results regarding the impact of policy uncertainty on bank risk. Bank risk is the dependent variable and is measured with five alternative proxies, including Z-score, σ(ROA), σ(NIM), LLP, and NPL, where higher values for each of these proxies represent higher bank risk and vice versa. Policy uncertainty is the main explanatory variable and is measured with the WUI (world uncertainty index) index of Ahir et al. (2018). Higher values for WUI represent higher policy uncertainty and vice versa. The others are bank-, banking industry-, and country-level control variables. The results are estimated with a pooled OLS estimator using heteroskedasticity robust standard errors. p-values are reported in parentheses. *** and ** represent statistical significance at the 1% and 5% levels, respectively.

, the WUI index enters positively and significantly in all five models. The results from Models 1 to 5 indicate that heightened policy uncertainty immediately boosts bank default risk, volatility in net interest income, volatility in bank overall operating income, expected loan losses, and realized loan losses, respectively.

The results are economically significant. For example, in Model (5), a one standard deviation increase in the WUI index (0.12) raises NPL by 0.81 (calculated as 6.768 × 0.12) compared to a mean NPL value of 6.42. This implies that a one standard deviation increase in policy uncertainty leads to an approximate 12.6% rise in non-performing loans.

Similarly, the EPU index, the alternative measure of policy uncertainty (in

Table 5. Impact of policy uncertainty on bank risk: main specifications.

Variables	Z-Score	σ(ROA)	σ(NIM)	LLP	NPL
	Model (1)	Model (2)	Model (3)	Model (4)	Model (5)
EPU	0.083 ***	0.060 ***	0.202 ***	0.082 ***	0.048
	(0.000)	(0.000)	(0.000)	(0.008)	(0.415)
Bank-level control variables
Bank size	−0.037 ***	−0.019 ***	−0.013 ***	−0.038 ***	−0.499 ***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Equity-to-total assets ratio	−0.030 ***	0.020 ***	0.017 ***	−0.014 ***	0.037 ***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Non-interest income-to-total revenue ratio	0.006 ***	0.003 ***	0.000	0.002 ***	0.025 ***
	(0.000)	(0.000)	(0.540)	(0.000)	(0.000)
Annual growth in bank gross loans	0.002 ***	0.001 ***	0.002 ***	−0.002 ***	−0.066 ***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Liquid assets-to-total assets ratio	0.004 ***	0.002 ***	0.002 ***	0.003 ***	0.017 ***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Cost-to-income ratio	0.013 ***	0.004 ***	0.002 ***	−0.014 ***	0.004 *
	(0.000)	(0.000)	(0.000)	(0.000)	(0.052)
Listed bank dummy	−0.071 ***	0.002	−0.007	0.103 ***	−0.503 ***
	(0.000)	(0.818)	(0.473)	(0.000)	(0.000)
State-owned bank dummy	0.178 ***	−0.074 ***	−0.055 ***	0.445 ***	4.272 ***
	(0.000)	(0.000)	(0.010)	(0.000)	(0.000)
Bank market power	0.004 **	0.002 **	0.000	0.008 ***	0.132 ***
	(0.012)	(0.013)	(0.936)	(0.000)	(0.000)
Banking industry-level control variables
Banking industry concentration	0.004 ***	−0.001 ***	0.001 ***	0.005 ***	−0.051 ***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Capital stringency index	−0.028 ***	−0.031 ***	−0.045 ***	−0.081 ***	−0.091 ***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.003)
Activity restrictions	−0.021 ***	−0.030 ***	−0.031 ***	−0.026 ***	0.434 ***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Country-level control variables
Developing countries dummy	0.774 ***	0.450 ***	0.481 ***	1.044 ***	−5.207 ***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
GDP growth rate	−0.005	−0.010 ***	−0.005 ***	−0.113 ***	−0.352 ***
	(0.102)	(0.000)	(0.002)	(0.000)	(0.000)
Inflation	0.014 ***	0.012 ***	0.018 ***	−0.003	−0.117 ***
	(0.000)	(0.000)	(0.000)	(0.454)	(0.000)
Law and order	−0.208 ***	−0.100 ***	−0.157 ***	−0.406 ***	−2.953 ***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Financial crisis dummy	0.119 ***	−0.014	0.022 **	0.048 *	0.499 ***
	(0.000)	(0.158)	(0.045)	(0.086)	(0.004)
Year FE	Yes	Yes	Yes	Yes	Yes
Constant	−3.269 ***	0.744 ***	0.494 ***	5.182 ***	30.164 ***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Observations	50,595	50,595	50,595	50,595	30,266
R-squared	0.249	0.403	0.394	0.226	0.248

This table reports results regarding the impact of policy uncertainty on bank risk. Bank risk is the dependent variable and is measured with five alternative proxies, including Z-score, σ(ROA), σ(NIM), LLP, and NPL, where higher values for each of these proxies represent higher bank risk and vice versa. Policy uncertainty is the main explanatory variable and is measured with the news-based EPU (economic policy uncertainty) index of Baker et al. (2016). Higher values for EPU represent higher policy uncertainty and vice versa. The others are bank-, banking industry-, and country-level control variables. The results are estimated with a pooled OLS estimator using heteroskedasticity robust standard errors. p-values are reported in parentheses. ***, **, and * represent statistical significance at the 1%, 5%, and 10% levels, respectively.

), also enters positively and significantly with bank risk in four Models, except when non-performing loans are used to measure bank risk. Together, these results are consistent with our prediction and confirm that higher uncertainty about government policy leads to higher bank risk.

The results for the control variables are also consistent with expectations, validating our empirical model. For example, negative coefficients of bank size indicate that large banks exhibit lower risk. The positive results of the non-interest income-to-total revenue ratio and growth in bank gross loans show that diversified and growing banks experience higher risks.

Similarly, higher levels of liquid assets are associated with increased risk. This aligns with findings that banks may hoard liquidity by curtailing lending (Acharya and Skeie 2011) and excessive liquid holdings driven by precautionary motives can exacerbate agency problems (Acharya et al. 2012), resulting in higher bank risks (Delis et al. 2014; Khan et al. 2017). Regarding operational efficiency, a higher cost-to-income ratio correlates with increased default risk, income volatility, and non-performing loans. Despite these risks, inefficient banks maintain lower loss provisions, suggesting imprudent behavior in managing expected credit losses.

State-owned banks exhibit higher default risk and loan losses but lower earnings volatility. This may stem from a mandate to lend to high-risk priority sectors and from government backing. Listed banks, conversely, demonstrate lower default risk, reduced income volatility, and fewer NPLs, while maintaining higher loss provisions. These findings suggest that market monitoring and investor pressure encourage prudent risk management and more robust provisioning (Samet et al. 2018; Tran et al. 2019).

The results for industry-level control variables show that strict financial regulations, such as stringent capital requirements and higher activity restrictions, are effective in controlling bank risk. These findings are consistent with the results of recent studies (Rahman et al. 2015; Ashraf et al. 2016b).

For country-level macroeconomic controls, the positive and significant coefficients of the developing countries dummy variable indicate that bank risk is higher in developing countries as compared to developed ones. The possible explanation for this finding is that banks in developed countries have better access to advanced risk management techniques, diversification opportunities, and human and capital resources. The negative coefficients of GDP growth show that bank risk is lower when the economy is in the growth phase of the business cycle. These findings are consistent with the studies that report that bank risk, especially non-performing loans and loan loss provisions, is pro-cyclical (Bikker and Metzemakers 2005; Laeven and Huizinga 2019): that is, it decreases when GDP growth is higher and vice versa.

Likewise, negative results for law and order suggest that a better institutional environment promotes bank stability by lowering bank risk and thus is beneficial.

Lastly, positive results for the financial crisis variable show that bank default risk, income volatility, and loan losses increase during the crisis situation.

5.3. Robustness Tests

We performed several robustness tests to further verify the main results. In this regard, we performed tests for endogeneity, added additional control variables to the main model, applied alternative estimation methods, and dropped countries with a higher number of observations. For brevity, we only report robustness tests for the WUI index.

5.3.1. Tests for Endogeneity

Endogeneity might be a potential concern with our results. Endogeneity may arise because of at least three reasons: reverse causality, measurement error, or omitted variables.

There might be reverse causality between policy uncertainty and bank risk. Bloom (2014) observes that policy uncertainty and domestic business cycles are countercyclical; that is, uncertainty goes down in booms and up in recessions. In a similar vein, if excessive loan losses start materializing on bank balance sheets, governments usually respond with policies to avoid a large-scale crisis. Such policy responses often entail uncertainty regarding the likeliest and best set of options available to be used by governments. To account for this reverse causality problem, we measure policy uncertainty together with all other control variables at the start of the year and re-estimate the results. As shown in

Table 6. Impact of policy uncertainty on bank risk: robustness tests for endogeneity.

Variables	Z-Score	σ(ROA)	σ(NIM)	LLP	NPL
	Model (1)	Model (2)	Model (3)	Model (4)	Model (5)
L.WUI	0.339 ***	0.103 ***	0.045 *	0.437 ***	5.759 ***
	(0.000)	(0.000)	(0.096)	(0.000)	(0.000)
Bank-level control variables
L.Bank size	−0.051 ***	−0.021 ***	−0.012 ***	−0.032 ***	−0.587 ***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
L.Equity-to-total assets ratio	−0.028 ***	0.020 ***	0.015 ***	−0.005 ***	0.003
	(0.000)	(0.000)	(0.000)	(0.000)	(0.698)
L.Non-interest income-to-total revenue ratio	0.008 ***	0.003 ***	−0.000 ***	0.003 ***	0.026 ***
	(0.000)	(0.000)	(0.006)	(0.000)	(0.000)
L.Annual growth in bank gross loans	0.002 ***	0.001 ***	0.001 ***	−0.001 ***	−0.059 ***
	(0.000)	(0.000)	(0.000)	(0.005)	(0.000)
L.Liquid assets-to-total assets ratio	0.002 ***	0.001 ***	0.002 ***	−0.001 ***	0.006 *
	(0.000)	(0.000)	(0.000)	(0.004)	(0.061)
L.Cost-to-income ratio	0.009 ***	0.003 ***	0.002 ***	−0.007 ***	−0.011 ***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
L.Listed bank dummy	−0.099 ***	−0.021 **	−0.029 ***	0.039	−0.729 ***
	(0.000)	(0.013)	(0.002)	(0.129)	(0.000)
L.State-owned bank dummy	0.226 ***	−0.025	−0.047 **	0.515 ***	4.374 ***
	(0.000)	(0.224)	(0.035)	(0.000)	(0.000)
L.Bank market power	0.006 ***	0.002 ***	−0.000	0.012 ***	0.158 ***
	(0.000)	(0.004)	(0.670)	(0.000)	(0.000)
Banking industry-level control variables
L.Banking industry concentration	0.002 ***	−0.002 ***	0.000	−0.000	−0.055 ***
	(0.000)	(0.000)	(0.536)	(0.985)	(0.000)
L.Capital stringency index	−0.018 ***	−0.024 ***	−0.026 ***	−0.070 ***	−0.044
	(0.000)	(0.000)	(0.000)	(0.000)	(0.133)
L.Activity restrictions	−0.014 ***	−0.028 ***	−0.035 ***	−0.039 ***	0.555 ***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Country-level control variables
L.Developing countries dummy	0.678 ***	0.412 ***	0.499 ***	1.225 ***	−5.878 ***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
L.GDP growth rate	−0.021 ***	−0.016 ***	−0.015 ***	−0.105 ***	−0.390 ***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
L.Inflation	0.025 ***	0.018 ***	0.023 ***	−0.004	0.098 ***
	(0.000)	(0.000)	(0.000)	(0.351)	(0.000)
L.Law and order	−0.205 ***	−0.090 ***	−0.146 ***	−0.363 ***	−2.860 ***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
L.Financial crisis dummy	0.148 ***	−0.015	0.014	0.151 ***	1.723 ***
	(0.000)	(0.104)	(0.176)	(0.000)	(0.000)
Year FE	Yes	Yes	Yes	Yes	Yes
Constant	−2.369 ***	1.061 ***	1.339 ***	4.822 ***	28.503 ***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Observations	42,482	42,482	42,482	42,482	25,289
R-squared	0.227	0.402	0.395	0.223	0.269

This table reports results regarding the impact of policy uncertainty on bank risk when all independent variables are lagged by one period to test the reverse causality problem. Bank risk is the dependent variable and is measured with five alternative proxies, including Z-score, σ(ROA), σ(NIM), LLP, and NPL, where higher values for each of these proxies represent higher bank risk and vice versa. Policy uncertainty is the main explanatory variable and is measured with the WUI (world uncertainty index) index of Ahir et al. (2018). Higher values of WUI represent higher policy uncertainty and vice versa. The others are bank-, banking industry-, and country-level control variables. The results are estimated with a pooled OLS estimator using heteroskedasticity robust standard errors. p-values are reported in parentheses. ***, **, and * represent statistical significance at the 1%, 5%, and 10% levels, respectively.

, the WUI index remains positive and significant, confirming the main results. One interesting observation is that the coefficients of lagged WUI are lower compared to the coefficients reported in Table 4, suggesting that the previous year’s uncertainty has a weaker effect on bank risk than the current year’s uncertainty.

Since we have used two alternative proxies of policy uncertainty that were constructed by different authors using the underlying data from different sources, the largely similar results for both proxies confirm that measurement error is less likely to be a concern regarding our findings.

Lastly, we have added a large number of bank-, banking industry-, and country-level control variables to the main model, which, to some extent, confirms that our results are not driven by some important omitted variables. However, to further account for omitted variable bias, we use two-stage instrumental variable analysis. A valid instrument should be relevant and exogenous: that is, the instrument should be directly correlated with policy uncertainty but only indirectly with bank risk through the channel of policy uncertainty. Following El Ghoul et al. (2021), we use the fractionalization index from the Database of Political Institutions (DPI2017) as the instrument for policy uncertainty.

The fractionalization index measures the probability that two deputies picked from a legislature at random would be from different political parties. This index is relevant because higher values of it represent a thin majority for any single party in the legislature and thus higher chances of different parties disagreeing on policy-related legislation (i.e., higher policy uncertainty). Aghion et al. (2004) find that legislative actions are more likely to be blocked with higher political fractionalization. On the other hand, the index is also exogenous because the distribution of members of political parties in a legislature is less likely to have any direct effect on bank risk.

In the first-stage regression, we regress the WUI index on the instrumental variable, together with all control variables in the model. In

Table 7. Impact of policy uncertainty on bank risk: robustness tests for endogeneity.

Variables	WUI	Z-Score	σ(ROA)	σ(NIM)	LLP	NPL
	First-Stage Regression	Second-Stage Regressions
	Model (1)	Model (2)	Model (3)	Model (4)	Model (5)	Model (6)
Fractionalization index	0.059 ***
	(0.000)
WUI_instrumented		1.549 **	4.043 ***	5.324 ***	8.738 ***	82.801 ***
		(0.028)	(0.000)	(0.000)	(0.000)	(0.000)
Bank-level control variables	Yes	Yes	Yes	Yes	Yes	Yes
Banking industry-level control variables	Yes	Yes	Yes	Yes	Yes	Yes
Country-level control variables	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes
Constant	0.181 ***	−3.346 ***	−0.037	−0.108	3.169 ***	6.114 ***
	(0.000)	(0.000)	(0.707)	(0.321)	(0.000)	(0.000)
Observations	50,221	50,221	50,221	50,221	50,221	29,959
R-squared	0.479	0.247	0.406	0.394	0.229	0.249

This table reports the results of the instrumental variable analysis regarding the impact of policy uncertainty on bank risk. The WUI (world uncertainty index) index of Ahir et al. (2018), which measures policy uncertainty, is instrumented with the fractionalization index from the Database of Political Institutions (DPI2017). Model (1) is the first-stage regression, where WUI is used as the dependent variable. Models (2) to (6) are second-stage regressions, where bank risk is the dependent variable, which is measured with five alternative proxies, including Z-score, σ(ROA), σ(NIM), LLP, and NPL. Higher values for each of these five proxies represent higher bank risk and vice versa. WUI_instrumented, which is the predicted values of the WUI index from the first-stage regression, is used as the main explanatory variable in second-stage regressions. Bank-, banking industry-, and country-level control variables are added to all models. The results are estimated with a pooled OLS estimator using heteroskedasticity robust standard errors. p-values are reported in parentheses. *** and ** represent statistical significance at the 1% and 5% levels, respectively.

, we observe that the fractionalization index enters positively and significantly with the WUI index. We rely on the F-test and the Kleibergen–Paap under-identification test to check the appropriateness of the instrument. The F-test in first-stage regression rejects the null hypothesis that the fractionalization index does not explain the variation in policy uncertainty, confirming that the instrument is relevant. Likewise, the Kleibergen–Paap rk LM statistic reports a zero p-value, showing that the model is identified, and the fractionalization index is an appropriate external instrument for policy uncertainty.

The fitted values of WUI from the first-stage regression are then used to represent instrumented policy uncertainty (i.e., WUI_instrumented) in the second-stage regressions. As shown in Table 7, instrumented WUI enters positively and significantly with all five proxies of bank risk.

Together, the results of these tests drive out the concern that the positive impact of policy uncertainty on bank risk found above is due to the endogeneity problem.

5.3.2. Political Risk Index from the ICRG Database as an Additional Control Variable

As an additional robustness check for omitted variable bias, we augment our model with the political risk index from the ICRG database. This index captures a broad spectrum of political risk factors, including government instability, internal and external conflicts, the nature of the political system (e.g., democracy versus autocracy), corruption levels, law and order conditions, and military involvement in politics. Models (1) to (5) in

Table 8. Impact of policy uncertainty on bank risk: robustness tests with an additional control variable and country fixed effects.

Variables	Z-Score	σ(ROA)	σ(NIM)	LLP	NPL	Z-Score	σ(ROA)	σ(NIM)	LLP	NPL
	Including the Political Risk Index as an Additional Control Variable					Including Country Fixed Effects
	Model (1)	Model (2)	Model (3)	Model (4)	Model (5)	Model (6)	Model (7)	Model (8)	Model (9)	Model (10)
WUI	0.404 ***	0.142 ***	0.252 ***	0.482 ***	3.992 ***	0.299 ***	0.050 **	0.118 ***	0.331 ***	4.327 ***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.034)	(0.000)	(0.000)	(0.000)
Bank-level control variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Banking industry-level control variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Country-level control variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Political risk	−0.033 ***	−0.005 ***	0.002 **	−0.034 ***	−0.409 ***
	(0.000)	(0.000)	(0.011)	(0.000)	(0.000)
Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Country FE						Yes	Yes	Yes	Yes	Yes
Constant	−0.752 ***	1.277 ***	1.053 ***	7.547 ***	55.393 ***	−2.845 ***	0.809 ***	0.669 ***	4.266 ***	1.735
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.291)
Observations	50,595	50,595	50,595	50,595	30,266	50,595	50,595	50,595	50,595	30,266
R-squared	0.261	0.404	0.390	0.233	0.296	0.278	0.429	0.424	0.242	0.371

This table reports results regarding the impact of policy uncertainty on bank risk, after including the political risk index from the ICRG database as an additional control variable, and country fixed effects. Bank risk is the dependent variable in all models of Panels 1 and 2 and is measured with five alternative proxies, including Z-score, σ(ROA), σ(NIM), LLP, and NPL, where higher values for each of these proxies represent higher bank risk and vice versa. Policy uncertainty is the main explanatory variable and is measured with the WUI (world uncertainty index) index of Ahir et al. (2018). Higher values for WUI represent higher policy uncertainty and vice versa. Bank-, banking industry-, and country-level control variables are included in all models. The results are estimated with the pooled OLS estimator using heteroskedasticity robust standard errors. p-values are reported in parentheses. *** and ** represent statistical significance at the 1% and 5% levels, respectively.

show that the WUI index remains positive and statistically significant even after controlling for political risk, indicating that our proxies for policy uncertainty are not merely reflecting general political risk but rather the effect of government economic policy uncertainty. The political risk variable enters with a negative sign, suggesting that lower political risk (as higher index values represent lower risk) reduces bank risk. In unreported results, we further control for individual sub-components of the political risk index, such as government instability, democratic accountability, internal conflict, external conflict, and corruption, one at a time, and find that the WUI index results remain largely unchanged.

5.3.3. Adding Country Fixed Effects

Previous research highlights the role of country-level formal institutions, such as legal origin and creditor rights, and informal institutions, such as national culture, in explaining cross-country variation in bank risk. Although our results are less likely to be substantially influenced by these factors, given the limited number of sample countries, we nonetheless conduct additional tests to address this concern. Because institutional variables tend to remain constant or change only gradually over time, we account for them by re-estimating our main model with country-fixed effects for brevity. Country-fixed effects largely capture all time-invariant factors that drive cross-country differences in bank risk. Models (6) to (10) in Table 8 show that the WUI index continues to enter positively and significantly after including country fixed effects, ruling out the possibility that WUI is merely reflecting institutional characteristics.

5.3.4. Alternative Estimation Methods

As an additional robustness check, we employ panel random-effects and fixed-effects models as alternative estimation approaches. We re-estimate all five specifications from Table 4 using these two estimators separately. As shown in

Table 9. Impact of policy uncertainty on bank risk: robustness tests using alternative estimation techniques.

Variables	Z-Score	σ(ROA)	σ(NIM)	LLP	NPL	Z-Score	σ(ROA)	σ(NIM)	LLP	NPL
	Panel Random-Effects Estimator					Panel Fixed-Effects Estimator
	Model (1)	Model (2)	Model (3)	Model (4)	Model (5)	Model (6)	Model (7)	Model (8)	Model (9)	Model (10)
WUI	0.385 ***	0.097 ***	0.177 ***	0.560 ***	4.885 ***	0.280 ***	0.050 **	0.126 ***	0.464 ***	4.015 ***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.012)	(0.000)	(0.000)	(0.000)
Bank-level control variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Banking industry-level control variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Country-level control variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Constant	−3.212 ***	0.971 ***	1.290 ***	5.565 ***	19.322 ***	−5.238 ***	0.788 ***	1.131 ***	4.424 ***	3.193
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.101)
Observations	50,595	50,595	50,595	50,595	30,266	50,595	50,595	50,595	50,595	30,266
R-squared						0.168	0.051	0.023	0.129	0.227
Banks	5138	5138	5138	5138	4443	5138	5138	5138	5138	4443

This table reports results regarding the impact of policy uncertainty on bank risk, estimated with panel random-effects and fixed-effects estimators. Bank risk is the dependent variable in all models and is measured with five alternative proxies, including Z-score, σ(ROA), σ(NIM), LLP, and NPL, where higher values for each of these proxies represent higher bank risk and vice versa. Policy uncertainty is the main explanatory variable and is measured with the WUI (world uncertainty index) index of Ahir et al. (2018). Higher values for WUI represent higher policy uncertainty and vice versa. Bank-, banking industry-, and country-level control variables are added to all models. p-values are reported in parentheses. *** and ** represent statistical significance at the 1% and 5% levels, respectively.

, the WUI index remains positive and statistically significant across all models, confirming that our findings are not driven by the choice of estimation technique.

5.3.5. Dropping Countries with a Higher Number of Observations

The sample distribution in Table 1 indicates that the number of bank observations is disproportionately higher for certain countries, such as Germany (23,442), Japan (8797), and Italy (4906), compared to others. To address potential bias arising from this distribution, we sequentially exclude all observations from Germany, Japan, and Italy and re-estimate the five specifications reported in Table 4. As shown in

Table 10. Impact of policy uncertainty on bank risk: robustness tests dropping countries with a higher number of observations.

Variables	Z-Score	σ(ROA)	σ(NIM)	LLP	NPL	Z-Score	σ(ROA)	σ(NIM)	LLP	NPL	Z-Score	σ(ROA)	σ(NIM)	LLP	NPL
	All Observations of Germany Dropped					All Observations of Germany and Japan Dropped					All Observations of Germany, Japan, and Italy Dropped
	Model (1)	Model (2)	Model (3)	Model (4)	Model (5)	Model (6)	Model (7)	Model (8)	Model (9)	Model (10)	Model (11)	Model (12)	Model (13)	Model (14)	Model (15)
WUI	0.810 ***	0.108 ***	0.160 ***	0.940 ***	7.797 ***	0.674 ***	0.126 **	0.284 ***	0.989 ***	6.530 ***	0.536 ***	0.077	0.393 ***	0.777 ***	2.923 ***
	(0.000)	(0.002)	(0.000)	(0.000)	(0.000)	(0.000)	(0.013)	(0.000)	(0.000)	(0.000)	(0.000)	(0.215)	(0.000)	(0.000)	(0.000)
Bank-level control variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Banking industry-level control variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Country-level control variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Constant	−2.595 ***	1.238 ***	1.403 ***	5.015 ***	29.794 ***	−2.692 ***	1.170 ***	1.397 ***	4.852 ***	28.582 ***	−3.216 ***	1.190 ***	1.344 ***	4.604 ***	13.664 ***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Observations	27,153	27,153	27,153	27,153	23,632	18,356	18,356	18,356	18,356	15,192	13,450	13,450	13,450	13,450	10,407
R-squared	0.301	0.403	0.387	0.210	0.256	0.293	0.357	0.341	0.187	0.281	0.334	0.348	0.320	0.196	0.204

This table reports results regarding the impact of policy uncertainty on bank risk after excluding all observations of Germany, Japan, and Italy, one by one, from the main sample. Bank risk is the dependent variable in all models and is measured with five alternative proxies, including Z-score, σ(ROA), σ(NIM), LLP, and NPL, where higher values for each of these proxies represent higher bank risk and vice versa. Policy uncertainty is the main explanatory variable and is measured with the WUI (world uncertainty index) index of Ahir et al. (2018). Higher values for WUI represent higher policy uncertainty and vice versa. Bank-, banking industry-, and country-level control variables are added to all models. The results are estimated with a pooled OLS estimator using heteroskedasticity robust standard errors. p-values are reported in parentheses. *** and ** represent statistical significance at the 1% and 5% levels, respectively.

, the WUI index remains positive and statistically significant across all models, confirming that our results are not driven by the overrepresentation of a few countries.

5.4. Impact of Policy Uncertainty on Bank Risk with a Lag

To test the second hypothesis, we re-estimate Equation (1) by lagging WUI and all associated control variables sequentially by one year, two years, three years, and four years. As reported in

Table 11. Impact of policy uncertainty on bank risk with a lag.

Variables	Z-Score	LLP	NPL	Z-Score	LLP	NPL	Z-Score	LLP	NPL	Z-Score	LLP	NPL
	WUI Lagged by One Year			WUI Lagged by Two Years			WUI Lagged by Three Years			WUI Lagged by Four Years
	Model (1)	Model (2)	Model (3)	Model (4)	Model (5)	Model (6)	Model (7)	Model (8)	Model (9)	Model (10)	Model (11)	Model (12)
L.WUI	0.339 ***	0.437 ***	5.759 ***
	(0.000)	(0.000)	(0.000)
L2.WUI				−0.023	−0.401 ***	4.280 ***
				(0.703)	(0.000)	(0.000)
L3.WUI							−0.454 ***	−0.468 ***	0.510
							(0.000)	(0.000)	(0.283)
L4.WUI										−0.254 ***	−0.618 ***	−0.857
										(0.002)	(0.000)	(0.112)
Bank-level control variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Banking industry-level control variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Country-level control variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Constant	−2.369 ***	4.822 ***	28.503 ***	−2.289 ***	4.909 ***	28.836 ***	−2.227 ***	4.533 ***	26.431 ***	−2.423 ***	4.473 ***	22.616 ***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Observations	42,482	42,482	25,289	37,487	37,487	22,934	32,494	32,494	20,293	28,414	28,414	18,057
R-squared	0.227	0.223	0.269	0.204	0.211	0.272	0.182	0.220	0.288	0.173	0.251	0.318

This table presents the results on the impact of policy uncertainty on bank risk with lags of two to four years. Bank risk serves as the dependent variable in all models and is measured using three alternative proxies, Z-score, LLP, and NPL, where higher values of each proxy indicate greater bank risk. Policy uncertainty is the main explanatory variable, captured by the World Uncertainty Index (WUI) developed by Ahir et al. (2018), with higher WUI values representing greater uncertainty. L.WUI, L2.WUI, L3.WUI, and L4.WUI denote the WUI index lagged by one, two, three, and four years, respectively. All models include bank-, banking industry-, and country-level control variables, which are lagged by the same period as the WUI index. p-values are reported in parentheses, and *** indicates statistical significance at the 1% level.

, the coefficients of WUI remain positive when lagged by one year. However, some coefficients switch to negative when WUI is lagged by two years. At a three-year lag, WUI becomes negative and statistically significant for bank default risk in Model (11) and expected loan losses in Model (14). The level of significance further increases when WUI is lagged by four years. These findings suggest that policy uncertainty reduces bank default risk and loan losses with a lag of two to four years.

We exclude σ(ROA) and σ(NIM) risk measures from the lagged effect analysis. As explained in Section 4, both σ(ROA) and σ(NIM) are computed using a three-year rolling window. This construction method tends to smooth short-term fluctuations in profitability and net interest margins (Delis et al. 2014), rendering these measures unsuitable for lagged effect analysis.

As a robustness check, we replace WUI with the EPU index and re-estimate all specifications in Table 11. In unreported results, we find that the coefficients of the EPU index turn negative for bank default risk and loan loss provisions when lagged by three and four years, results that are consistent with the findings for the WUI index.

6. Conclusions

This paper investigates the impact of government EPU on bank risk, distinguishing between short- and long-term effects. We hypothesize that heightened EPU increases bank risk in the short run by raising borrowers’ default probabilities while reducing risk over time as banks adopt more conservative current lending strategies. Using bank-level data from 22 countries over the period 1998–2017, we find strong evidence supporting both hypotheses: EPU exerts an immediate positive effect on bank risk and a negative effect with a lag of two to four years. These results remain robust to endogeneity concerns, alternative measures of EPU and bank risk, variations in sample composition, and different estimation techniques.

Our findings have critical implications for policy-makers and financial institutions. First, the short-term increase in bank credit risks following policy uncertainty spikes heightens financial instability, suggesting that lower policy uncertainty is preferable for maintaining a stable financial system. Consequently, banks’ stress testing and risk management frameworks should explicitly incorporate the effects of varying policy uncertainty levels. Furthermore, monetary and fiscal policy-making institutions, such as central banks, must strive to design a policy environment characterized by minimum ambiguity. Crucially, it must be recognized that policy uncertainty plays a dual role in shaping bank risk, amplifying risk in the immediate term while promoting mitigation with a lag. This dynamic underscores the importance of integrating both the short-term amplification and the long-term stabilizing effects of uncertainty into institutional risk management and regulatory frameworks.

Future research could extend this analysis by exploring the impact of economic policy uncertainty on other dimensions of banking, such as profitability, lending behavior, business models, dividend policies, and earnings quality, while explicitly distinguishing between short- and long-term effects. For example, higher EPU can reduce current-period profits by lowering lending activity and increasing loan loss provisions and non-performing loans. At the same time, reduced lending today diminishes future interest income, thereby weakening profitability in subsequent periods. Unlike existing studies that primarily emphasize immediate outcomes, such investigations would offer a more comprehensive understanding of the dynamic influence of uncertainty on banking.

Additionally, country-level case studies could enrich the discussion by examining how institutional and regulatory environments shape the short- versus long-term consequences of policy uncertainty for bank risk.

One limitation of our study is that we analyze lagged effects using panel models. Future research could employ more advanced time-series techniques to provide a deeper assessment of these lagged relationships. Moreover, access to higher-frequency data, such as quarterly accounting information, could enable more granular and insightful analysis of the timing and persistence of policy uncertainty effects. Extending the sample beyond 2017 offers another valuable avenue for future research, as it would capture the effects of more recent economic and geopolitical developments, including the COVID-19 pandemic, the Russia–Ukraine conflict, and changes in U.S. trade tariffs.