Week 45 Day 2: Historical vs. Forward-Looking: Choosing Your Inputs

Historical (1926-2024): stocks averaged approximately 10%/year. Bonds averaged approximately 5%/year. Forward-looking (based on current P/E ratios and bond yields): stocks 7-8%/year. Bonds 3-4%/year. The difference seems small but compounds enormously: $1,000,000 at 10% for 30 years = $17,400,000. At 7% for 30 years = $7,600,000. Using historical returns in your simulation could make a risky plan look safe.

Why forward-looking estimates differ from historical: (1) Stock market valuations. The Shiller CAPE (cyclically adjusted P/E) ratio is currently approximately 33-36. Historically, when CAPE > 30, subsequent 10-year stock returns averaged 3-5%/year. When CAPE < 15, subsequent returns averaged 12-15%/year. High valuations today imply lower future returns. (2) Bond yields. Bond returns are approximately predictable: your expected return over the next 10 years is approximately equal to today's yield. Current 10-year Treasury yield: approximately 4.0-4.5%. This is better than the 1.5-2% yields of 2020-2021 but lower than the 5-7% yields of the 1990s. (3) Inflation expectations. The Fed targets 2% inflation. Recent inflation has been 3-4%. Higher inflation erodes real returns. Building your simulation inputs: Conservative (recommended for planning): Stocks: 7% nominal (5% real). Bonds: 4% nominal (2% real). Inflation: 2.5%-3%. Moderate: Stocks: 8.5% nominal (6% real). Bonds: 4.5% nominal (2% real). Inflation: 2.5%. Optimistic (use for upside scenarios, not base planning): Stocks: 10% nominal (7.5% real). Bonds: 5% nominal (2.5% real). Inflation: 2.5%. Run your simulation with ALL THREE input sets. If your plan succeeds at conservative inputs, you have a robust plan. If it only succeeds at optimistic inputs, you are dependent on favorable markets. (4) Do not chase precision. The difference between 7% and 8% expected return is unknowable in advance. The point of using conservative inputs is not to predict the future accurately but to build a plan that works even if the future is below-average.

The expected equity return can be decomposed using the Gordon growth model: E[R] = Dividend yield + Earnings growth + Valuation change. Current S&P 500 dividend yield: approximately 1.3%. Historical real earnings growth: approximately 2-3%/year. Expected valuation change: if the CAPE ratio reverts from approximately 34 to the long-run average of approximately 17 over 20 years, this would subtract approximately 3.5%/year from returns. If CAPE stays elevated (a 'new normal'), valuation change is 0%. This gives a range of expected real equity returns: Bear case (CAPE reversion): 1.3% + 2.0% - 3.5% = approximately 0%/year real. Base case (partial reversion): 1.3% + 2.5% - 1.0% = approximately 3%/year real (6% nominal). Bull case (no reversion): 1.3% + 3.0% + 0% = approximately 4.3%/year real (7% nominal). The historical average of approximately 7% real return was achieved partly because of valuation expansion (CAPE increased from approximately 10 in the early 20th century to approximately 34 today). This one-time expansion boosted historical returns above sustainable levels. Starting from today's elevated valuations, repeating history would require CAPE to reach approximately 100 -- implausible. Research supports using forward-looking estimates: Bogle (2009) advocated the 'sources of return' approach (yield + growth +/- valuation change). Arnott, Bernstein, and Hall (2003) showed that equity returns in the following decade were strongly correlated (r approximately 0.5-0.6) with starting CAPE and dividend yield. Pfau (2013) showed that using forward-looking return estimates (based on current valuations) in Monte Carlo simulations produced success rates approximately 10-20 percentage points lower than using historical averages -- a materially different planning outcome.