Jonas Mohamed Osman Abdelghafour, known as Yonas Osman
LanguageEnglishDeutschSvenskaالعربية

Quantitative Methods

Economic Scenario Generators for Pricing, Capital and ALM

An economic scenario generator produces internally consistent simulated paths for interest rates, inflation, equity returns, credit spreads and exchange rates. It underpins market-risk capital, asset and liability management and the valuation of options and guarantees.

By Jonas Mohamed Osman Abdelghafour, known as Yonas Osman · Published · Reviewed · 4 min read

Monte Carlo simulation paths and probability density used in quantitative risk modelling — Economic Scenario Generators for Pricing, Capital and ALM, analysis by Jonas Mohamed Osman Abdelghafour, known as Yonas Osman
Figure 1. Schematic view of simulated paths and the resulting distribution referenced in this analysis.

Executive summary

  • Real-world calibration supports capital and planning; risk-neutral calibration supports market-consistent valuation. The two are not interchangeable.
  • Interest-rate models must produce plausible curve shapes and dynamics, not only plausible short rates.
  • Dependence between economic variables drives tail results and is the hardest element to calibrate credibly.
  • Validation should test distributional properties, dynamic behaviour and the reproduction of known relationships, not only marginal fit.
  • Governance requires documented calibration targets, version control and independent review of each recalibration.

Two purposes, two calibrations

A real-world scenario set aims to represent the plausible distribution of future economic outcomes. It is used for capital modelling, strategic asset allocation, ALM and planning, where the question is what could happen and how likely it is.

A risk-neutral scenario set aims to reproduce observed market prices when used for valuation. It is used for pricing options and guarantees and for market-consistent liability valuation. Its paths are not intended to be realistic descriptions of the future and should never be presented as forecasts.

Confusing the two is a recurring and consequential error: using risk-neutral paths for capital produces distributions that do not reflect real-world probability, and using real-world paths for market-consistent valuation produces values that do not reconcile to observable prices.

The component models

  • Nominal interest rates: multi-factor short-rate or forward-rate models capable of generating level, slope and curvature movements, with an explicit position on negative rates.
  • Inflation: modelled jointly with nominal rates so that implied real rates remain economically sensible.
  • Equity: returns with time-varying volatility and fat tails, linked to the rate and inflation environment.
  • Credit: spread dynamics plus transition and default behaviour, with spreads that widen in adverse states.
  • Exchange rates: consistent with interest-rate differentials across the modelled economies.
  • Property and alternative assets: usually modelled more simply, with explicit acknowledgement of smoothing in the underlying index data.

Internal consistency is the defining requirement. Each component can be individually well specified while the combination produces economically incoherent states — high inflation with falling nominal rates and tightening credit spreads, for instance. Reviewing joint states, not only marginals, is essential.

Dependence and tail behaviour

Correlation estimated over benign periods understates the co-movement that occurs in stress. Equity and credit, in particular, exhibit stronger joint tail behaviour than a multivariate normal specification implies. Copula structures with tail dependence, or regime-switching specifications, address this more honestly than adjusting a correlation matrix upward by judgement.

  • Test whether the generated scenarios reproduce historically observed joint extremes, not only historical correlations.
  • Report the diversification benefit implied by the dependence structure as a separate figure so its magnitude is visible.
  • Run sensitivity analysis on the dependence assumption; if capital moves substantially, the assumption deserves board attention.

Calibration and validation

  1. Document the calibration targets: which market data, which historical window, which statistics are being matched and why.
  2. Test marginal distributions against the targets and against long-run historical experience.
  3. Test dynamic properties: mean reversion speed, volatility clustering, autocorrelation and curve-shape evolution.
  4. Test economic relationships: real rates, term premia and equity risk premia should remain within defensible ranges across the scenario set.
  5. Check convergence: confirm that the number of scenarios is sufficient for stability in the tail statistics actually used.
  6. Version-control every scenario set and record which business decisions used which version.

Practical example

An insurer generates 10,000 real-world scenarios for a five-year capital projection. Reviewing joint states reveals that in 3 per cent of paths, inflation exceeds 8 per cent while ten-year nominal yields remain below 2 per cent for the full period — a combination with no historical precedent and no coherent economic mechanism in the model.

The marginal distributions for inflation and for yields were each individually reasonable. The defect was in the joint specification. Because such paths fall in the region that drives the capital result, correcting the dependence structure changed the capital figure materially. Reviewing only marginal calibration would never have identified it.

Limitations and caveats

  • Calibration to a historical window embeds the characteristics of that window, including any structural regime that has since changed.
  • Tail dependence is estimated from few joint extreme observations and carries wide uncertainty.
  • Model risk in the interest-rate specification propagates to every dependent variable.
  • Simulation error at extreme percentiles requires large scenario counts that may be computationally constrained.
  • Scenario sets are conditional on the model structure; a different structure can produce materially different tails from the same data.

Conclusion

An economic scenario generator is one of the most influential models in an insurer or bank, because so many other results depend on it.

Its validation should focus on joint behaviour, dynamic properties and economic coherence — the elements that determine tail results — rather than on marginal fit, which is comparatively easy to achieve and comparatively uninformative.

References

Author bio

Jonas Mohamed Osman Abdelghafour, known as Yonas Osman, actuary and financial risk professional

Jonas Mohamed Osman Abdelghafour, known as Yonas Osman is an actuary, FRM and financial risk professional specialising in banking, insurance, model risk, capital modelling and quantitative risk management.