Staff Insights

Key Insights

• No single geopolitical or policy uncertainty indicator captures all financial stability risks, with economic context and forecast horizon of key concern. Looking one quarter ahead, a geopolitical risk indicator can act as a useful near-term discriminant of systemic banking crises, assuming not missing crises is your policy preference. Assuming policy preference indifference, a more standard indicator of financial market stress remains preferred. 

• Looking four quarters ahead, sharp spikes in other global economic policy and trade policy uncertainty indicators signal elevated downside risk to output growth, rendering them potentially valuable additions to financial stability toolkits.

• Global policy uncertainty lowers Irish equity prices and consumer confidence within one quarter, with corporate credit weakening thereafter, while house prices and financial stress remain broadly stable in our estimates.

Introduction

Recent periods of heightened uncertainty (whether geopolitical, financial or policy-driven) have been associated with sharp movements in macro-financial variables, and occasionally the outbreak of systemic financial distress (see Beck et al. (2025), Hodula et al. (2024), Caldara and Iacoviello (2022), Baker et al. (2016)).^[1] For a small, highly open economy such as Ireland’s, external uncertainty shocks originating for example from global geopolitical events or abrupt shifts in international economic policy can propagate rapidly.

Understanding how uncertainty transmits to economic outcomes is crucial for financial stability risk assessment. The literature identifies several key transmission channels. Bloom (2009) demonstrates that heightened uncertainty triggers “wait-and-see” behaviour, causing firms and households to delay irreversible investment and consumption decisions until information improves. This real options channel can lead to sharp declines in economic activity even without direct financial disruption. Beck et al. (2025) show that geopolitical shocks can also affect banks’ financial soundness directly through credit and market risk exposures. For small, open economies, these effects may be amplified through trade linkages and multinational business decision-making, where policy uncertainty in major economies can rapidly affect domestic investment and employment (Rice, 2023).^[2]

This Insight examines the information content of several widely used global uncertainty indicators in assessing and forecasting economic and financial risk. Specifically, we ask:

1. Do these indicators yield any early warning signals of near-term systemic banking crises?

2. Do they contribute to deteriorating macroeconomic growth-related tail risk forecasts?

3. Through what macro-financial channels do they affect a small, open economy such as Ireland?

As global uncertainty indicators, we consider the Geopolitical Risk (GPR) index of Caldara and Iacoviello (2022), the (Global) Economic Policy Uncertainty (EPU) index of Baker, Bloom and Davis (2016), the Trade Policy Uncertainty (TPU) index of Caldara and Iacoviello (2022), and the Common Global Geopolitical Volatility (COVOL) indicator of Engle and Campos-Marins (2023). The first three are text-based measures that capture uncertainty through automated analysis of news coverage and policy discourse, while COVOL is a model-based indicator derived from a factor model of volatility across asset classes.

The Country Level Index of Financial Stress (CLIFS, ECB), integrated with our in-house Irish Composite Stress Indicator (ICSI),^[3] and the Chicago Board Options Exchange Volatility Index (VIX, a measure of expected global/S&P 500 equity volatility) are established real-time risk indicators often used to measure and signal potential financial market turbulence.^[4] These form a benchmark against which global uncertainty indicators can be compared. An overview of each of these indicators is outlined in Table 1. The behaviour of these indicators depends critically on whether uncertainty shocks are temporary or persistent, policy or market-driven, and whether they occur during periods of existing financial vulnerability.

Financial, Geopolitical and Policy Uncertainty Indicators

Table 1: Source and description of indicators considered in this Insight

Source: Central Bank of Ireland
Note: Authors’ summary of the effects documented in the papers cited. Irish modified domestic demand (MDD) replaces Irish GDP in all models considered in this Insight.

Are these indicators useful as early-warning indicators of bank crises?

Assessment of crisis-signalling power via ROC analysis

How concerned should we be about the near-term ramifications of an indicator’s upward spike? For instance, can such spikes be suggestive of an imminent systemic banking crisis? ^[5] To answer these questions, we compare the baseline crisis-signalling power of the various indicators relative to proven benchmark indicators such as CLIFS. We evaluate their performance from a 1 quarter ahead proximity to banking crises perspective and follow the standard Receiver-Operating-Characteristic (ROC) curve approach used in the early-warning literature (see Parla (2021), Du Plessis (2024), Huynh and Uebelmesser (2024) and Alonso-Alvarez and Molina (2023)). The ROC curve plots the trade-off between correctly predicting a banking crisis (true positive rate or sensitivity) and being overconfident in predicting a banking crisis when none occurred (false positive rate or 1 – specificity) across a range of possible crisis-signalling thresholds.^[6]

The Area-Under-the-ROC (AUROC) measures an indicator’s predictive abilities: a value of 1 indicates perfect discrimination between crisis and tranquil periods, while a value of 0.5 implies no information content. Our analysis uses a pooled logit specification across 27 countries in the OECD database with crises periods defined by the Laeven and Valencia (2020) database of financial crises.^[7] The CLIFS index serves as a benchmark stress measure against which the crisis signalling ability of the other indices may be gauged for all countries apart from Ireland, for which we use ICSI (see also Parla, 2021). We refer to this “combined” indicator as CLIFS/ICSI.

Comparative performance of indicators

Results (see Table 2) show that the benchmark CLIFS/ICSI index yields the highest general discriminatory power with an AUROC of 0.78, confirming it as one of the most effective single-variable predictors of bank crises. The GPR follows with an AUROC score of 0.74, also suggesting robust crisis signalling power. By contrast, the other indicators appear to generate more noisy signals.

Policy preferences and signal-to-noise trade-offs

The shape of the ROC curve (see Figure 1) yields additional valuable insight. The CLIFS/ICSI curve is consistently above the diagonal, thereby highlighting the extent to which crisis signals generated by the model outweigh noisy or false signals.^[8] Policymaker preferences with respect to true and false positives might now come into focus. Assume there exists a policymaker or national authority whose primary preference is to miss very few, if any, actual crises. For example, their preference might be to not “miss” more than 1 in 10 actual crises (sensitivity = 90%). Further assume this requirement outweighs the reputational effect of occasionally “crying wolf” by predicting crises that never actually materialise. Conditional upon this minimum true positive threshold, the indicators might then be ranked according to their false positive score, with lower values of the latter preferred. On this basis, GPR ranks first because it generates fewer false positives whilst meeting the 90% true positive requirement.

Ranking of indicators according to extent of crisis signals generated

Table 2: General noise to signal ranking and under an assumed policy preference

EWS Properties of Geopolitical and Policy Uncertainty Indicators

Source: Central Bank of Ireland
Note: The 90% true positive threshold is arbitrarily assigned by the authors and is for exposition purposes only. Policymakers might have different preferences which would have to be modelled on a case-by-case basis.

Implications for financial stability monitoring

The GPR index’s performance, second only to CLIFS /ICSI, suggests that major geopolitical shocks could presage imminent banking sector difficulties, with the effects dependent on the scale of the sector’s credit and market risk exposures to the specific shock. GPR spikes, by their nature, are diffuse. By this we mean that the risks they track stem from a variety of sources and may become manifest across multiple indirect channels affecting the financial system. The resilience of the financial system, investor and consumer confidence and country level effects also matter. We explore these channels in more detail below. The weaker performance of TPU and COVOL is significant. Relative to the other indicators, they appear less useful in predicting financial crises over the sample and may be more useful as coincident indicators of heightened financial stress.

Are they macroeconomic vulnerability indicators?

Assessing downside risk through Growth-at-Risk analysis

To answer this question, we next consider how uncertainty shapes future macroeconomic forecasts using a Growth-at-Risk (GaR) framework. Adapted from Adrian et al. (2019) and O’Brien and Wosser (2021), our approach links uncertainty indicators to the conditional full distribution of GDP growth (Modified Domestic Demand in the case of Ireland) one year ahead.^[9] In terms of the model outputs, the focus is on the coefficients reported at lower quantiles, as these inform adverse growth forecasts. Indicators with negative coefficients are suggestive of greater downside risk as the indicator increases in value. We find divergent behaviour.

Divergent effects on economic tail risk

GPR, COVOL, and VIX have positive coefficients at low quantiles, counter-intuitively associating higher values with improved left tail forecasts.^[10] The positive GPR low percentile coefficients likely reflect a “policy-response” channel: major geopolitical and geoeconomic events trigger coordinated public intervention supporting near-term forecasts, hence masking the underlying shock. This timing mismatch limits GPR’s usefulness for forward-looking tail risk assessment. Additionally, the GPR’s construction (see Table 1) may not reliably capture the effect of specific shocks upon geographically disparate countries. The observed patterns suggest that GPR, COVOL, and VIX are not reliable forward-looking indicators of Irish macroeconomic tail risk. The CLIFS/ICSI variable, though primarily a financial-stress measure, reflects a mildly negative association at lower percentiles, reinforcing its persistent utility as a forward-looking risk indicator.

Policy uncertainty as leading indicator

EPU and TPU show statistically significant negative coefficients at the 5th-10th percentiles. Rising policy or trade uncertainty thus worsens downside prospects for Irish economic growth, plausibly through delaying investment and hiring decisions.^[11] Therefore, policy-related uncertainty indices EPU and TPU may contain useful leading information about Irish economic vulnerability, while purely geopolitical or volatility measures may not. One interpretation is that financial markets quickly price in exogenous geopolitical events, limiting their predictive content for future activity, whereas policy uncertainty affects the decision horizons of firms and households more persistently.

An example of how these results affect MDD forecast distributions is outlined in Figure 3, where the inclusion of EPU in the GaR model results in increased tail risk to MDD growth at a forecast horizon of T+4Q ahead.

How does policy uncertainty affect Ireland?

Transmission mechanisms in the Irish macro-financial environment

Above, we established the usefulness of the EPU index for demand-based Irish macroeconomic tail risk forecasts. How does such uncertainty transmit to real outcomes through the macro-financial environment? While internal factors are at play as well, for a small, open economy like Ireland policy uncertainty is driven largely by developments in external markets and by multinational business whose decision-making bodies are often based elsewhere. Such global policy uncertainty can be thought of as exogenous, something that arrives from outside the Irish economy and is beyond the control of domestic policymakers.^[12] To trace how policy uncertainty propagates through the Irish macro-financial environment, we use a Vector Autoregression with Exogenous Variables (VARX) model,^[13] which is designed to capture how multiple economic variables influence each other over time: changes in one variable feed back into others, creating complex dynamics that unfold over time. The VARX variant allows EPU to affect the system without being itself affected by other variables.

Short-term financial market responses

We include EPU in levels rather than as a shock, reflecting our interest in how the prevailing uncertainty environment affects the macro-financial system over time. Figure 4 summarises our main results. A two-standard-deviation EPU increase, which roughly corresponds to the cumulative change observed between late 2024 and early 2025, reduces real equity prices by a modest but meaningful amount, with the largest decline occurring within the first three months. This is consistent with real options theory, where heightened policy uncertainty causes firms and consumers to delay irreversible decisions until information improves (Bloom, 2009). The equity market response reflects investor concerns about future corporate earnings given anticipated declines in investment and consumption. Consumer confidence follows a similar pattern, declining within a quarter as households’ subjective assessments of economic conditions deteriorate

Credit conditions

While financial markets react within a quarter, bank lending to firms shows a more delayed response. This delay reflects the time required for the initial shock to translate into business decisions. When firms observe rising policy uncertainty, they become more cautious, delaying major capital expenditure projects until they have greater clarity about the policy environment.

The model shows that corporate financing costs may ease in the short run following an EPU shock, even as credit volumes fall, consistent with the interpretation that policy uncertainty operates as a demand-side effect. Central banks may respond to economic shocks by easing monetary policy, which helps support financial conditions and keeps borrowing costs down even as the underlying economic shock dampens demand.^[14] However, this demand-driven pattern might reflect the specific sample period in which monetary policy remained accommodative, and the banking sector’s health was generally strong. Supply-side credit tightening could dominate in other scenarios. The model finds that policy uncertainty increases have negligible effects on real house prices and the overall financial stress index over the sample period.

Implications for financial stability

Our VAR results portray a transmission mechanism which operates primarily through the real economy, consistent with wait-and-see behaviour: firms delay investment decisions and consumers postpone major purchases. This anticipated slowdown in economic activity explains both the equity market decline (as investors revise down earnings expectations) and the fall in credit demand.

The analysis offers several important lessons for financial stability. The immediate transmission to equity prices and confidence, followed by lagged effects on corporate credit, shows that the Irish financial system is not insulated from global uncertainty shocks. The demand-driven nature of the transmission suggests that the primary concern is not a sudden credit crunch or financial panic, but rather a gradual erosion of growth prospects among firms.

Our findings warrant careful interpretation, as they reflect average historical responses and do not imply that future shocks will necessarily trigger similar responses. Our modelling choices are estimated on historical relationships, and if the structure of the Irish economy or financial system were to change substantially, the historical relationships might become obsolete. For example, the 2025 H2 Central Bank’s Financial Stability Review (PDF 1.41MB) highlighted the disconnect between low measured stock market volatility and high policy uncertainty in 2025.

Conclusions

As part of its remit to monitor and maintain the stability of Ireland’s financial system, the Central Bank pays close attention to emerging external risks and geopolitical developments. Recent unprecedented events such as the Covid-19 outbreak and wars in Ukraine and the Middle East disturbed financial markets and raised significant economic concerns for governments, firms, and households. Relatively new indicators such as GPR, EPU, COVOL and TPU track these concerns.

The Central Bank continues to examine a broad spectrum of indicators to understand their effectiveness as tools for monitoring and managing systemic risk to the Irish economy and their most informative context. In this context, we investigate the properties of the new indicators and benchmark their performance relative to more-established risk indices such as the CLIFS index.

All the indices we examine appear to reflect major global shocks to varying degrees.^[15] We find that, on a prima facie basis, the CLIFS/ICSI remains, overall, the most informative early-warning indicator of systemic banking crises. The other more globally focused variables add context but display weaker discriminative power. Within a Growth-at-Risk (GAR) framework, EPU and TPU emerge as statistically significant predictors of deteriorating left-tail forecasts, whereas GPR, COVOL and VIX provide little or even counter-intuitive information. Finally, we find that, when global policy uncertainty rises, it is typically followed by a decrease in stock market capitalisation, consumer confidence, and corporate credit demand based on Irish data over 2003 to early 2025.

Our findings support the use of multiple indicators when assessing policy uncertainty-related financial stability risks. CLIFS/ICSI remains the cornerstone real-time indicator for domestic financial-system stress, but GPR is also promising in this regard. EPU and TPU serve a complementary role in that they provide valuable forward-looking information about macroeconomic tail risk and transmission to real-economy variables. COVOL, and VIX provide useful context for understanding global market sentiment but should not be relied upon for forecasting Irish systemic risk or macroeconomic downturns.

Appendix

For the purposes of this analysis, we aggregate values of the indicators into quarterly averages to better compare their properties and to better fit the frequencies used in several of the Central Bank of Ireland’s risk models.

Irish Composite Stress Index (ICSI)

Developed by Parla (2021), the Irish Composite Stress Index synthesises information from a range of domestic financial market segments, including money-market spreads, sovereign and bank bond yields, credit default swap premia, equity market volatility and liquidity indicators, into a single monthly measure of financial stress. It employs a variance-equal-weighting scheme like that used in the ECB’s Composite Indicator of Systemic Stress, standardising component series and aggregating them into sub-indices for banking, securities and foreign-exchange markets. The ICSI rises when tensions in any of these markets increase, offering a concise summary of systemic stress conditions in Ireland’s financial system. Because of its proven utility in risk models, it represents a benchmark against which other geopolitical or uncertainty measures may be compared.

Geopolitical Risk Index (GPR)

Iacoviello and Caldara’s GPR index quantifies the intensity of geopolitical tensions as measured by automated text searches of major international newspapers for terms related to military conflicts, terrorism and diplomatic crises. The monthly index extends back to 1985 and captures broad swings in perceived geopolitical threat. The GPR has been shown to spike around major military events such as the Gulf Wars, 9/11, and the invasion of Ukraine. Its relevance for Ireland arises through global confidence and trade channels whereby elevated geopolitical risk can dampen global investment and financial flows, indirectly affecting Irish output and asset markets.

COVOL Indicator (Common Global Geopolitical Volatility)

Engle and Campos-Marins introduced the COVOL indicator to measure the common latent component of geopolitical volatility extracted from a large panel of national GPR indices. It applies a dynamic conditional correlation model to capture co-movements in geopolitical risk across countries. In principle, COVOL isolates the truly global element of geopolitical stress, filtering out idiosyncratic local events. For a small, open economy such as Ireland’s, sensitivity to global risk sentiment implies that COVOL may better capture the exogenous component of uncertainty shocks relevant to Irish financial conditions.

Economic Policy Uncertainty (EPU) Index

The EPU index of Baker, Bloom and Davis is based on newspaper coverage frequency of terms related to “economic,” “policy,” and “uncertainty.” It also incorporates measures of fiscal policy disagreement and forecasts dispersion. The index has been interpreted as reflecting uncertainty about future economic policy settings involving tax, regulation, and macroeconomic management. It has been shown to explain fluctuations in investment and hiring (Bloom 2009). For Ireland, the EPU serves as a proxy for external policy ambiguity, especially in major trading partners such as the US and the UK.

Trade Policy Uncertainty (TPU) Index

Caldara and Iacoviello (2022) extend the EPU methodology to a trade-specific domain by counting references to tariff and trade-agreement uncertainty in leading international newspapers. The TPU surged during the US–China trade dispute and provides a direct measure of uncertainty regarding cross-border trade policy. Given Ireland’s heavy export dependence and integration into multinational production chains, the TPU is potentially a highly relevant external uncertainty shock for Irish growth and corporate sentiment.

VIX Index

The VIX index, derived from implied volatility on S&P 500 options, is a benchmark measure of global financial market risk aversion. Though not a geopolitical indicator per se, it captures real-time expectations of equity-market volatility and often reacts immediately to news events. For Ireland, the VIX serves as a useful “financial conditions” reference against which the predictive content of the other indices can be compared.

Table A: Summary statistics of data utilised in this Insight

Source: CLIFS from ECB data portal. ICSI from Central Bank calculations. GPR, EPU and TPU from www.policyuncertainty.com COVOL from New York University (STERN) Volatility Laboratory (V-LAB). VIX index is sourced via Federal Reserve Dataset.
Note: Any errors or omissions are our own.

Visual assessment of indicator responsiveness to major shocks

To gauge how well the indicators capture several historically significant events, we visually inspect their time-series behaviour in the periods circa eight representative landmark global and Irish financial events: A) the dot-com bust (2000–02); B) the 9/11 attacks (2001); C) the collapse of Lehman Brothers (2008); D) Ireland’s sovereign-bond downgrade (2010); E) the Brexit referendum (2016); F) the global outbreak of COVID-19 (2020); G) Russia’s invasion of Ukraine (2022); and H) the US Presidential election (2024).

A review of these series suggests the indicators’ behaviour around major events varies in terms of magnitude and timing. The ICSI peaks sharply during Ireland’s sovereign debt crisis in 2010-11 and again during the onset of COVID-19, reflecting domestic funding-market stress. The VIX responds promptly to almost every global shock, consistent with its interpretation as a barometer of global financial turmoil.

Among the policy uncertainty measures, the EPU appears to exhibit the strongest event-tracking behaviour. It records notable jumps around five of the eight benchmark events examined, involving events B, C, E, F and H. The GPR and TPU show weaker correspondence. The COVOL indicator generally moves in the same direction as GPR but with smaller amplitude, suggesting that while it captures common volatility, its smoothing properties dampen short-lived spikes.

In early 2025, the TPU index surged to an unprecedented level (10 standard deviations above its historical mean value), coinciding with new US tariff announcements. This extreme outlier signals significant perceived global disruption from trade conflict and possibly a regime shift, not just a periodic development. While useful for monitoring extreme market sentiment, such extreme value shifts complicate statistical modelling, as its non-stationary variance can result in implausible risk-model forecasts.

This exercise suggests that whereas visual correlations can identify broad co-movements, they should be interpreted cautiously. No single indicator perfectly tracks all major financial shocks (see Table B). The EPU’s comparatively high hit-rate may also suggest how policy uncertainty influences global macro-financial cycles, whereas the subdued responses of GPR and TPU imply that not all geopolitical and trade disturbances translate directly into global, or Irish specifically, financial stress.

Summarising the behaviour of indicators circa recent major financial events

Table B: Accounting for spikes in indicators close to major events

Index Tracking at Key Milestones

Source: Central Bank of Ireland
Note: In absence of well-established thresholds, we assign a score of 1 to an indicator if it visually appears to spike upwards around the time of the financial shock involved. Selected events include A) Dotcom bubble market correction, B) 9/11 attack, C) Lehman Brothers failure, D) Irish sovereign bond downgrade, E) Brexit referendum, F) Covid outbreak, G) Ukraine invasion and H) US 2024 presidential election.

Endnotes

1. Authors: Michael Wosser, Economist, Macro-Financial Division, and Emil Bandoni, Economist, Macro-Financial Division. We wish to thank Mark Cassidy, Niamh Halissey, Martin O’Brien, Maria Woods, and Angelos Athanasopoulos, as well as seminar participants at the Central Bank, for helpful comments and suggestions. All views expressed in this Insight are those of the authors alone and do not necessarily represent the views of Central Bank of Ireland. ↑
2. For Ireland, the transmission of global shocks has likely evolved after the Global Financial Crisis. The Irish banking sector’s funding profile shifted substantially from wholesale to deposit-based funding, potentially altering the speed and channels through which external uncertainty affects domestic financial stability. While our panel analysis captures average effects across countries and time periods, these structural changes warrant consideration when interpreting results for Ireland specifically. ↑
3. The Irish Composite Stress Index (ICSI), adapted by Parla (2021), replaces Irish CLIFS data in our panel for reasons set forth in Parla (2021). The remaining countries in the panel reflect prevailing financial conditions via their respective CLIFS indices. Henceforth we refer to CLIFS/ICSI in combination, to reflect this risk-modelling preference, one which the Central Bank has adopted since 2020. ICSI incorporates information from money, sovereign bonds, equity, banking and foreign exchange markets by using the time-varying correlation-based methodology proposed by Hollo et al. (2012). ↑
4. We considered, but ultimately decided against, utilising the Irish variants of the EPU index (see Rice (2023)) in this analysis for two primary reasons. Firstly, the dataset underpinning two of the risk models described in this Insight involves a cross-country panel of OECD-sourced data. As such, the authors’ view is that the use of a common global EPU variable is better suited to our purposes. Secondly, in this Insight’s final model, we envisage an EPU spike as an exogenous global shock affecting policy uncertainty more broadly rather than a country-specific shock, wherein the case for exogeneity is less compelling. Our approach therefore ensures that significant EPU increases are treated on a consistent basis across each model. ↑
5. In the Appendices, we chart and summarise the behaviour of these indicators over time, paying close attention to periods during which significant financial disruption was observed. ↑
6. This analysis is exploratory in nature and simply intended to motivate the rest of the analysis that follows. Our goal at this stage is to highlight whether, on their own, the various indicators examined help to discriminate crisis periods from non-crisis periods and if increases in the indicators are associated with higher probabilities of systemic banking crises in the near term. Additional control variables are not included. ↑
7. See the Appendices for more details on the panel size and the countries comprising the panel. ↑
8. Sensitivity and specificity involve setting a threshold for an indicator; if exceeded, it is assumed to signal systemic banking crisis. The resulting signals are compared against a crisis database to determine true and false positive rates. This process is repeated to find the most informative threshold values. ↑
9. For details on the modelling approach refer to O'Brien and Wosser (2021). ↑
10. Robustness checks, where GDP replaces MDD for IE, were undertaken without changing any of our core findings or interpretation. ↑
11. It is important to qualify our TPU results with caution. The TPU spikes markedly on only two prior occasions in the timeframe we consider (during the 2018-2019 US trade dispute with China and in 2025Q2). The 2025 spike is so large that it may dilute the effects of more modest TPU index increases in our models. Refer to the commentary relating to the time series behaviour of the indices in the Appendices for more details. ↑
12. Granger causality tests reveal that EPU is unaffected by Irish macro-financial variables. ↑
13. The model includes four lags of each endogenous variable. Quarterly frequency data from 2003 to early 2025 are used. Following the variables selection from the ESRB report on geoeconomic fragmentation, the endogenous variables in the model are real equity prices, real house prices, a financial stress index aggregating multiple signals of market tension, consumer confidence reflecting household expectations, inflation, credit to non-financial corporations, and corporate financing costs. The model is estimated using Bayesian methods. ↑
14. Because the VAR does not include the ECB policy rate, we cannot disentangle monetary policy effects from movements in bank lending spreads. ↑
15. Refer to the Appendices for further details. ↑