Narratives and risks: a quantitative toolkit to evaluate scenarios applied to the Middle East conflict

Prepared by Niccolò Battistini, Mohammed Chahad and Luís Fonseca

Published as part of the ECB Economic Bulletin, Issue 7/2026.

This box introduces a quantitative framework to evaluate the alternative scenarios around the Eurosystem/ECB staff baseline projections, focusing on assessing the plausibility of their economic impact. As is typical in a policy context, the baseline projections are assumed to represent the most likely outcome within the set of possible future outcomes, conditional on the current state of the economy and a set of technical assumptions. To highlight the uncertainty around the baseline, these projections are often accompanied by alternative scenarios which explore the macroeconomic implications of policy-relevant economic narratives (European Central Bank, 2025). As stressed in Cordova-Pozo and Rouwette (2023), however, the literature assessing scenario relevance remains limited.^[1] This box proposes a methodology aimed at partially closing this gap by evaluating scenarios, with a focus on the plausibility of their macroeconomic outcomes compared with empirical patterns.^[2]

The first step in assessing scenario plausibility is to construct joint predictive distributions capturing temporal and cross-variable patterns. The objective is to obtain a joint density for a set of macroeconomic variables – in this application, real GDP growth and HICP inflation centred around the baseline and accounting for both temporal and cross-variable correlations. This is achieved by estimating dynamic relationships between variables, with two complementary approaches building on univariate distributions.^[3] First, a model-based approach derives univariate conditional distributions via Bayesian quantile regressions and combines them into a joint density by applying the chain rule of probability. Second, a copula-based approach constructs joint densities by combining univariate marginal distributions with historical patterns. In this application, these marginals are derived from two sources − (i) historical data and (ii) a survey-based distribution (the ECB Survey of Professional Forecasters) − incorporating professional forecasters’ forward-looking beliefs regarding future economic developments.^[4] The resulting joint densities are centred so that their mode corresponds to the baseline projections. These centred joint distributions are computed over 20-year rolling windows starting from the early 1980s in order to judge scenarios against the historical period most favourable to them. By computing the density over rolling windows, the framework captures time-varying relationships between variables. This is crucial, since the simulated scenarios might reflect various temporal and cross-variable dynamics that might differ from any specific historical pattern.

The second step assesses the empirical plausibility of scenario outcomes by computing how likely each scenario is relative to the baseline. In line with Doan et al. (1984), plausibility can be gauged by assessing how unusual a scenario path is relative to the most likely trajectory.^[5] The empirical plausibility index is obtained by first computing, for each historical subsample, the likelihood of the joint paths of growth and inflation under a given scenario relative to the likelihood of the baseline paths.^[6] The final index corresponds to the highest value across subsamples for a given scenario. It ranges from 0 – when the economic outcomes of the scenario have virtually no historical precedent – to 100 – when the scenario paths are as plausible as the baseline. Importantly, the score evaluates whether the joint co-movements of the outcome variables of a specific scenario – rather than the events underlying its narrative (e.g. specific disruptions in energy markets or a change to a given tariff regime) – have materialised in the past. Consequently, it can be applied to a broad set of scenario narratives, including those that may have never actually occurred.

Applying the framework to the Middle East conflict scenarios in the September 2026 ECB staff projections shows that the severe scenario has a lower empirical plausibility than the milder or adverse scenarios. Chart A, panel a) reports the paths of real GDP growth and HICP inflation as outlined in the Middle East conflict scenarios published in Box 4 of the September 2026 ECB staff macroeconomic projections for the euro area. These scenarios differ in their assumptions about energy commodity prices and financial uncertainty, as well as about the strength of indirect and second-round effects on inflation. They span a large set of economic outcomes, with cumulative absolute deviations from the baseline projections ranging from 0.2 to 1.2 percentage points for real GDP growth and from 1.0 to 4.3 percentage points for HICP inflation. Panel b) presents the plausibility index for each scenario, revealing that the milder scenario demonstrates high plausibility, with an index ranging between 82 and 96 across different approaches. For the adverse scenario, the range is between 56 and 85, while the severe scenario is assessed as considerably less plausible. Despite its lower plausibility, it remains important to analyse this scenario in order to cover a wide range of conflict outcomes relevant to policy decisions. The plausibility assessment is robust across the different frameworks used to derive the joint densities, although this need not be the case by design given the substantial methodological differences in those frameworks.

Chart A

Empirical plausibility of baseline and alternative scenarios

a) Baseline and alternative scenarios

(year-on-year change, percentages)

b) Empirical plausibility index across selected scenarios

(index, baseline = 100)

The empirical plausibility of scenarios goes beyond a simple assessment of their severity, as it also reflects both temporal and cross-variable correlations. The nexus between plausibility and severity can be illustrated by comparing assessments of different scenarios with similar severity. For instance, the Middle East adverse scenario exhibits a level of severity – measured as the cumulative absolute deviation from baseline values – similar to an ad hoc hypothetical flat-path scenario in which GDP growth and HICP inflation remain constant at their baseline averages over a three-year horizon. As shown in Table A, however, accounting for both temporal and cross-variable correlations reveals that the adverse scenario outcomes are far more plausible than those in the flat-path scenario, with empirical plausibility scores of 56 and almost 0 respectively. The comparison also shows that omitting either or both dependency dimensions results in a partial, often misleading, evaluation of the plausibility of the scenario outcomes. This is evident, for instance, for the hypothetical flat-path scenario: ignoring temporal or cross-variable dependencies yields high empirical plausibility estimates that can scarcely be corrected by any simple aggregation of these partial assessments, such as averaging.

In conclusion, the proposed quantitative framework to evaluate scenarios provides a measure of the empirical plausibility of scenario outcomes. It does not, however, assess the economic plausibility of narrative events or their probability of materialisation, as these assessments remain in the domain of policy judgement. Well-founded scenarios may receive low empirical plausibility scores if they generate combinations of growth and inflation outcomes with limited historical precedent (e.g. lockdown measures during the COVID-19 pandemic). Conversely, a novel scenario that exhibits standard macroeconomic co-movements may legitimately receive a high empirical plausibility score. In all cases, the metric should be interpreted with caution alongside the qualitative scenario narrative and relevant expert and policy judgement. Lastly, it should also be noted that the proposed methodology is not able to disentangle the contributions of key drivers to the empirical plausibility score, such as the shocks underlying the scenario, transmission channels and model misspecification.

References

Adrian, T., Giannone, D., Luciani, M. and West, M. (2025), “Scenario Synthesis and Macroeconomic Risk”, IMF Working Papers, No 2025/105, International Monetary Fund.

Cordova-Pozo, K. and Rouwette, E. (2023), “Types of scenario planning and their effectiveness: A review of reviews”, Futures, Vol. 149, 103153.

Doan, Y., Litterman, R. and Sims, C. (1984), “Forecasting and conditional projection using realistic prior distributions”, Econometric Reviews, Vol. 3, Issue 1, pp. 1-100.

European Central Bank (2025), “A strategic view on the economic and inflation environment in the euro area”, Occasional Paper Series, No 371, ECB.

Legal Disclaimer:

EIN Presswire provides this news content "as is" without warranty of any kind. We do not accept any responsibility or liability for the accuracy, content, images, videos, licenses, completeness, legality, or reliability of the information contained in this article. If you have any complaints or copyright issues related to this article, kindly contact the author above.