Market Simulations & Financial Planning | #411 (John Yang)

Episode Summary

Executive Summary: Episode 411 examines how PWL estimates expected returns and simulates portfolio outcomes, then tests a student-built upgrade to its Monte Carlo engine. The new approach uses a T-copula plus empirical tails/EVT to better match real return shapes and downside risk, while preserving co-movement. It modestly changes retirement spending outcomes, but meaningfully affects long-term wealth and estate projections.

Main Topics: Expected return estimation methodology (Priority: 5/5): Ben explains PWL’s blended framework: market-implied returns, long-run historical data, and asset-specific weighting for equities and bonds, plus factor premiums and tax-aware return composition. Geometric vs. arithmetic means in planning (Priority: 5/5): Brayden clarifies why projections with constant returns need geometric means, while simulated return paths require arithmetic means plus proper portfolio volatility/correlation adjustments. Why a new simulation method was needed (Priority: 5/5): The team wanted to improve beyond Gaussian Monte Carlo because normal distributions miss fat tails, skew, volatility clustering, and crisis co-movement—especially important for planning and alternatives. Columbia student project and John Yang interview (Priority: 5/5): John describes the synthetic-data framework: marginal asset distributions from historical data, T-copula dependence, and EVT/GPD tails to generate realistic market paths under forward-looking assumptions. Impact on financial planning outputs (Priority: 4/5): Brayden compares the new model with the old Gaussian approach and finds only small changes to spending success rates, but larger changes to wealth outcome distributions and estate planning assumptions. Why clients choose AUM advice (Priority: 4/5): A detailed client testimonial highlights reasons to hire an advisor: planning depth, cross-border complexity, estate security, cognitive-decline safeguards, expertise, accountability, and mental relief. Future model improvements (Priority: 4/5): The discussion ends with possible next steps: dynamic correlations, volatility clustering via GARCH-like methods, and alternative simulation approaches such as GANs, plus better proxies for private/alternative assets.

Key Arguments: PWL’s current expected-return framework blends market-implied signals with long-run historical data to balance current valuations and long-term evidence. A simple Gaussian Monte Carlo model is too smooth; it understates left-tail risk, fat tails, and crisis co-movement that matter for real financial planning. Using geometric returns alone can understate the benefit of diversification; converting to arithmetic means and back captures portfolio-level diversification benefits more accurately. The new simulation method improves the realism of distribution shape and downside risk without sacrificing correlation fidelity. For spending-focused plans, the new model does not dramatically change advice because success-rate and sustainable-spending outputs move only modestly. For estate- or legacy-focused plans, the improved shape matters more because it reduces unrealistic upper-tail runaway outcomes and changes mean/median wealth forecasts. The client testimonial argues that advisor value often comes from planning, coordination, and peace of mind—not just investment management. Trusted contacts and outside oversight can matter materially when cognitive decline or unexpected life events threaten financial decision-making.

Data Points: Episode number: 411 - Rational Reminder episode focused on expected returns and simulation modeling. Stock expected return weighting: 75% historical / 25% market-based - PWL’s equity expected return blend uses long-run historical returns plus valuation-implied return. Bond expected return weighting: 25% historical / 75% market-based - PWL gives bonds more weight on current yield to maturity because yields are more predictive. Simulation runs uploaded to planning software: 1,000 runs - PWL uploads simulated data into Conquest Planning for projections. Return-path dataset size mentioned later: about 4 million data points - Brayden describes the scale of simulated data sent to planning software. Planning success-rate change from new model: about +2% to +3% - Compared with the Gaussian baseline, the improved model modestly increases success rates. Sustainable-spending change: about 1% to 2% - Spending recommendations move only slightly under the new simulation method. Median final net worth change: about +$500,000 - The improved model raises the median wealth outcome in long-term projections. Mean final net worth change: about -$1,000,000 - The improved model reduces the mean because it reins in unrealistic upper-tail runaway outcomes. Model horizon: 80 years - John’s team generated 1,000 simulated return paths covering 80 years each. Historical data coverage for DMS: since 1900 - PWL uses Dimson-Marsh-Staunton long-run historical data as its equilibrium anchor. Index-data history used in the project: back to 2003 - The student team initially worked with monthly index data limited to a shorter sample period. Cognitive-decline study sample: Health and Retirement Study; age 50+ - Referenced research on awareness of cognitive decline and wealth losses. Finding on wealth losses: losses concentrated in highest wealth quartiles - The cited 2024 study found unaware respondents experienced more wealth loss, especially among wealthier households.

Pivotal Quotes: "The T copula plus empirical EVT framework is a stronger simulation baseline than the Gaussian model because it better reproduces historical shape and downside risk while maintaining comparable co-movement." — John Yang: John’s summary of the student project’s main model improvement. "We’re not forcing it into any of the shape. It’s the same as running an optimizer." — John Yang: Explaining why the empirical/tail-aware model better reflects real return distributions. "The main takeaway from this is that forward-looking assumptions are starting points when you're applying for wealth. To use those assumptions in planning, they had to be turned into paths." — John Yang: Closing reflection on why simulation methodology matters for financial planning.

Implications: The episode suggests better simulation methods can improve realism without overturning typical retirement spending advice, but they matter more for wealth-transfer, alternative assets, and concentrated positions. It also reinforces that advisor value often lies in planning, oversight, and behavioral support.