Forecasting approaches
- Statistical: historical means, shrinkage estimators (blend sample estimates with a target), time-series models.
- Discounted cash flow: yields for bonds; dividend models for equities.
- Risk premium (building blocks): add premiums to a risk-free rate.
- Financial equilibrium: international CAPM, Singer-Terhaar.
- Surveys and judgment.
Fixed income
Expected return over a horizon ≈
- yield income (coupon and rolldown),
- ± change in yields × duration,
- − credit losses,
- ± currency changes.
Building blocks: short-term default-free rate + term premium + credit premium + liquidity premium. Term premiums are higher when inflation uncertainty is high; credit premiums vary with the cycle.
Emerging market debt adds sovereign and currency risk; analyse fiscal and external balances, reserves and politics.
Equities
Grinold-Kroner model:
E(R) ≈ D/P − %ΔS + i + g + %ΔP/E
Three parts: income (D/P − %ΔS), nominal earnings growth (i + g) and repricing (%ΔP/E).
Singer-Terhaar (international CAPM with segmentation):
RP_i = φ × [ρ(i,GM) × σi × SR_GM] + (1 − φ) × [σi × SR_GM] + illiquidity premium
where φ is the degree of integration. A fully integrated market’s premium depends on its correlation with the global market; a segmented one’s on its total volatility.
Real estate
E(R) ≈ cap rate + NOI growth − expected change in cap rate. Unsmooth appraisal data before estimating volatility and correlations.
Currencies
Combine views on trade (current account), capital flows, interest rate differentials and PPP. Persistent current account deficits require capital inflows; when financing dries up, currencies adjust sharply.
Volatility and correlations
- Sample covariance matrix: noisy with many assets.
- Factor models: impose structure, reduce estimation error.
- Shrinkage: blend sample estimates with a target.
- Smoothed returns: unsmooth first.
- Time-varying volatility: ARCH/GARCH models.
Exam focus
Apply Grinold-Kroner and building blocks, use Singer-Terhaar, forecast real estate and currency returns, and choose methods for covariance estimation.