Executive Summary
This paper reviews the main approaches to Commercial Real Estate (CRE) loan risk assessment and suggests why cash-flow simulation provides the most robust framework for risk modelling.
While expert judgement cannot produce quantitative measures such as Probability of Default (PD) or Loss Given Default (LGD), it remains valuable as a qualitative overlay to model outputs.
The central finding is that traditional regression and scorecard models are unreliable for CRE loans. Even where large volumes of consistent default and loss data exist (rare in practice), regression models can only describe how loans behaved in the last downturn. Because future crises differ materially from past ones, such models provide weak guidance for forward-looking risk management.
The alternative approach has been to supplement expert judgement with a discounted cashflow model to assess the ability of the asset to support loan repayments and refinance at term. This approach is also challenged by its inability to provide a sound basis to understand either the probability of default or estimate losses in default.
By contrast, cash-flow simulation explicitly incorporates macroeconomic uncertainty and property-level dynamics, modelling loan outcomes across thousands of economic and market scenarios. This method reflects real-world lease structures, tenant risk, and property market cycles, while generating robust quantitative measures (PD, LGD, Expected Loss, capital requirements, etc.).
“Regression models tell you how loans performed in the last crisis. Cash-flow simulation tells you how they might behave in the next one.”
Scorecards and Regression Models
Scorecards assign scores to key loan and property characteristics (e.g., LTV, DSCR, lease length, tenant quality), weight them, and map results to ratings or PDs. Regression analysis calibrates these scores against historic defaults and losses.
However, this framework relies on assumptions that rarely hold for CRE lending:
- Large, consistent historical datasets
- Homogeneous loan structures
- Stable loan terms over time
- Defaults driven mainly by credit factors rather than market cycles
- Linear relationships between risk factors and losses
In reality, CRE lending features bespoke structures, sparse and inconsistent loss data, shifting loan terms, and outcomes dominated by market cycles. Many debt funds have been established in recent years and thus do not possess loss data over several recessions from which to build robust models. The history of CRE loan losses is almost entirely crisis-driven, not credit-factor driven.
Conclusion
Scorecards and regression models are not fit for purpose in CRE loan risk modelling.
Expert Judgement
Most lenders, particularly for larger CRE loans, rely heavily on expert judgement supported by ratio analysis (LTV, DSCR, exit yield, debt yield) based on a Discounted Cashflow model. There is no methodology to establish what each of these cases should be. Thus this approach is qualitative, often formalized through credit committees.
Strengths
- Incorporates soft factors (borrower integrity, management quality, property viability)
- Enables knowledge transfer and due diligence
- Anecdotally more effective at ranking loans than scorecards
Weaknesses
- Cannot generate quantitative risk metrics (PD, LGD, EL)
- Heavily influenced by institutional culture and approval pressures
- Arbitrary calibration of ratings
- Historically poor track record during crises (1992/93, 2007/08)
Cash-Flow Simulation Modelling
Cash-flow simulation treats CRE loans as asset-backed exposures whose repayment depends on uncertain property income and collateral value. It projects forward NOI and collateral performance under many macroeconomic and property market scenarios, using Monte Carlo simulation to generate thousands of potential loan outcomes.
Key features
- Models NOI based on rent rolls, lease expiries, tenant default probabilities, costs, inflation, and market rent assumptions.
- Simulates collateral value and refinancing conditions at maturity.
- Applies default tests across scenarios to estimate PD.
- Calculates losses in defaulting scenarios to estimate LGD and Maximum Probable Loss (MPL).
Unlike regression, this method does not assume homogeneity, linearity, or repeatable cycles. It directly incorporates market volatility in property values, rents, and interest rates; loan-specific structures (amortizing, interest-only, senior/junior tranches); and correlations between macroeconomic variables and real estate performance.
Conclusion
For CRE lending, cash-flow simulation is the most reliable risk modelling approach. It delivers consistent, quantitative, and forward-looking measures across different loan structures and market conditions - supporting PD, LGD, Expected Loss and regulatory capital requirements that regression and expert judgement cannot.