The forward-looking dimension of IFRS 9 is, in principle, what made the standard a meaningful improvement over the incurred-loss model that preceded it. In practice, it is also where most ECL implementations carry the largest unaddressed risk. The PD/LGD/EAD architecture can be sound, the staging logic can be defensible, and the data can be clean — and the model can still produce an output that does not credibly reflect the macroeconomic environment it is meant to anticipate.
The reason is usually not a failure of intent. It is that the macroeconomic scenarios feeding the model were calibrated against the wrong reference world. A bank operating in the UAE, Saudi Arabia, Bahrain, or Qatar is exposed to a set of economic variables that bear only partial resemblance to the OECD-weighted scenario libraries that many off-the-shelf forecasting services produce. The institution ends up running a model that is technically compliant, statistically credible against historical data, and quietly disconnected from the macroeconomic conditions it is actually exposed to.
What IFRS 9 actually asks for
The forward-looking requirement sits in paragraphs B5.5.49 through B5.5.54 of IFRS 9. The institution must consider “reasonable and supportable information that is available without undue cost or effort” — which is the standard’s deliberate hedge against requiring perfect foresight — and must “consider information about past events, current conditions and reasonable and supportable forecasts of future economic conditions.”
Paragraph B5.5.42 adds the multi-scenario obligation. The ECL must reflect a probability-weighted outcome across scenarios. The standard does not specify how many scenarios; institutions have settled on three (base, upside, downside) as a working convention, though five-scenario and continuous-distribution approaches are also defensible and in some cases preferred for portfolios with non-linear loss profiles.
What the standard does not do is tell the institution which variables to use, how to forecast them, or how to translate macro inputs into PD and LGD adjustments. That latitude is where most of the practical work — and most of the practical risk — lives.
Why generic scenarios understate GCC cyclicality
A set of variables that matters for portfolio losses in a Western European bank — unemployment, residential real estate price index, sovereign yield curve — is not the wrong set for a GCC bank, but it is an incomplete set. The variables that actually move credit risk in the region overlap with, but are not the same as, the OECD canon.
Crude oil price is the most obvious. The transmission from Brent to GCC credit risk is not direct on a one-quarter basis, but on an eighteen- to twenty-four-month basis it shows up in corporate top-line revenue, contract pipelines for construction and services, government project disbursements, and ultimately in default rates on the corporate and SME book. A scenario library that treats oil as a global growth variable rather than as a regional revenue driver materially understates the through-cycle volatility of GCC credit risk.
Real estate has the second-order role. Dubai residential and commercial price indices, Abu Dhabi office occupancy, the Saudi residential market, and the developing markets in Bahrain, Qatar, Kuwait and Oman each have different cycles, different liquidity characteristics, and different lender exposure profiles. Treating “GCC real estate” as one variable hides material divergence; treating the UAE as a single market hides the divergence between Dubai’s developer-led cycle and Abu Dhabi’s more sovereign-anchored one.
Hospitality and tourism matters for any bank with material exposure to Dubai, Doha, or increasingly the Saudi giga-project hospitality pipeline. These are sector-specific variables that the institution’s own RAM or sector-classification system will surface as a concentration, but that off-the-shelf macro libraries will not include as scenario inputs.
Government project pipelines — Vision 2030 disbursements in Saudi, the UAE federal spend trajectory, Qatar’s post-2022 infrastructure programme — drive the corporate and project finance books in a way that is not captured by GDP forecasts alone. A 2% GDP forecast can be consistent with a 30% swing in disbursement against the pipeline; the credit risk transmission runs through disbursement, not GDP.
The point is structural. The GCC economic environment is materially driven by variables that are either absent from, or under-weighted in, generic global scenario libraries. An institution relying on the off-the-shelf scenario set is, in effect, running an IFRS 9 model calibrated to a different economy than the one it lends in.
Scenario weighting and the supervisor’s question
The second area where forward-looking calibration tends to break down is scenario weighting. The standard requires probability weights to be defensible. In practice, weights of 50/25/25 (base/upside/downside) or 60/20/20 are the conventional defaults, often unchanged across reporting periods.
The supervisor’s question — increasingly common in CBUAE thematic reviews and analogous reviews at SAMA and CBB — is not whether the weights are reasonable in the abstract, but whether they have been updated, and on what basis, since the prior reporting period. A scenario weighting that does not move when the macroeconomic environment moves is a weighting the institution has stopped genuinely thinking about. A scenario weighting that moves at every reporting period without a documented rationale for each move is one that suggests the institution is managing the ECL output rather than calibrating the model.
The defensible middle ground is a scenario weighting framework with explicit triggers for review — material moves in oil, in regional yield curves, in real estate price indices, in government-spend forecasts — and a documented governance process for adjusting weights in response. The point is not that the weights should be volatile. It is that the process for setting them should be visible to a reviewer.
NGFS and the climate dimension
The Network for Greening the Financial System scenarios have been adopted, in varying forms, by central banks across Europe and parts of Asia. CBUAE has indicated supervisory interest in climate scenario integration into IFRS 9, though the regulatory expectation is still being developed at the time of writing. For GCC institutions specifically, the climate transmission to credit risk runs partly through the global oil price trajectory (the NGFS “delayed transition” scenarios produce materially different oil paths than the “current policies” scenarios), partly through physical risk to specific sectors (real estate exposure to extreme heat events, infrastructure exposure to coastal water-level changes), and partly through transition risk to specific obligors (hydrocarbon-linked corporates with concentrated business models).
The integration of NGFS into IFRS 9 is not a near-term regulatory requirement for most GCC institutions. It is, increasingly, a near-term audit committee question. The institutions that have begun building the linkage between climate scenario libraries and ECL scenario weights are doing so partly in anticipation of supervisory direction and partly because the linkage is becoming part of the broader Pillar 2 conversation.
A closing observation
The forward-looking element of IFRS 9 was always going to be the part of the standard that distinguished a thoughtful implementation from a routine one. A decade in, the institutions whose IFRS 9 models hold up best under supervisory review and audit dialogue are not necessarily the ones with the most sophisticated statistical machinery. They are the ones whose scenario library is calibrated to the economy they actually lend in, whose weights move when the environment moves, and whose documentation makes the link from scenario to overlay to portfolio output retrievable on a working day.
The variables that matter for the GCC are knowable. The scenario weights can be made defensible. The model does not need to be rebuilt to get either of those right — but the scenario architecture sitting on top of it often does.