SPY 2027: A Regime-Conditional Monte Carlo

Why Regime Conditioning Matters

Standard Monte Carlo projections for equity indices assume stationary return distributions — a single mean and volatility calibrated to the full historical sample. This produces symmetric fan charts that ignore what the macro environment is actually doing. Markets don't draw returns from one distribution. They oscillate between distinct states with different risk/reward characteristics.

We identify these states using the rolling z-score of net Federal Reserve liquidity (WALCL - RRPONTSYD - WTREGEN). When the z-score drops below -0.5, the system is in Tightening. Above +0.5, Easing. Between, Neutral. Each regime has its own return distribution, and transitions between regimes follow a Markov process with empirically estimated probabilities.

The chart below maps a decade of SPY price history against these regime classifications. The colored bands show which regime was active, and the return annotation shows how SPY performed during each period.

What the Regimes Tell Us

Three patterns emerge from the regime history:

• Tightening (red bands) — counterintuitively the strongest regime in this sample. Annualized return of 16.9% with the lowest volatility (14.5%). One possible explanation: during active liquidity contraction, much of the bad news is already priced, and markets may be anticipating policy reversal. However, this interpretation is not conclusive — the result could also reflect the specific composition of Tightening periods in our 10-year window. Persistence is extremely high (99.3% daily) — once in Tightening, the system tends to stay.
• Neutral (amber bands) — the weakest and most volatile regime (-16.6% annualized, 23.0% vol). This extreme negative return is partly a sample artifact: the COVID crash and several other sharp drawdowns fall within Neutral classification, which captures transition periods where directional conviction is lowest. The true unconditional Neutral return is likely less negative, but the elevated volatility is a robust finding.
• Easing (green bands) — strong returns (19.7%) but elevated volatility (24.9%). Liquidity injection periods coincide with strong rallies but also with the volatility that accompanies major policy shifts.

The current state is Tightening — the favorable regime in this sample's return/vol profile. But the risk is regime transition. The z-score at -0.64 places us in Tightening, but macro stress (recession probability at 31%, yield curve at +0.48%) is increasing the probability of transition to Neutral.

The Forward Projection

The fan chart below projects 5,000 simulated SPY paths through end of 2027. At each trading day, the simulation: (1) draws a regime transition from the Markov matrix, and (2) draws a daily return from the regime-conditional distribution. Paths compound multiplicatively.

But this isn't a vanilla regime model. We enhance the simulation with three macro overlays that adjust the return distribution based on the current economic environment — each one pulling from live data on the mplot platform.

Enhancement Decomposition

The waterfall above quantifies the impact of each overlay on the median projection:

1. Recession probability overlay — at 31%, this activates a downward mean adjustment (-4.4% annualized). The adjustment scales linearly from zero at P=15% to -15% at P=70%, so we're roughly 30% through the adjustment range. This also slightly increases volatility via a recession-conditional vol scaling factor.

2. CAPE mean-reversion drift — we estimate the current Shiller CAPE at ~38.0 by scaling the last available reading (30.8 from September 2023) for S&P 500 price appreciation and approximate earnings growth since then. This estimate aligns with independent sources (multpl.com reports ~38). At 38.0 vs. a long-run mean of 17.4, the implied return headwind is -0.94% annualized (dampened by 0.3x). The underlying CAPE data is stale, but the price-scaling estimation produces a reasonable proxy.

3. Conditional transition matrix — the yield curve spread (+0.48%) and recession probability together define a stress factor that modifies the Markov transition probabilities. Under current conditions, the daily probability of leaving the favorable Tightening regime roughly doubles (0.6% → 1.3%). Over 467 trading days, this compounds to a meaningfully higher probability of spending time in the adverse Neutral regime — though the Neutral regime's extreme negative return (-16.6%) likely overstates the actual risk due to sample composition effects noted above.

The net effect: the base regime model projects a median return of +26%, reflecting the historically favorable Tightening parameters. The enhanced model tempers this to +7.6%, with the probability of loss more than doubling from 17% to 40%. The true outcome likely falls somewhere between these two estimates — the base model ignores macro headwinds, while the enhanced model may overweight them due to the Neutral regime's sample-dependent severity.

Methodology

1. Regime identification. Rolling 252-day z-score of net liquidity. Z < -0.5 = Tightening, Z > 0.5 = Easing, else Neutral. Thresholds are symmetric and fixed — not optimized to the sample.
2. Parameter estimation. Regime-conditional mean and volatility of daily SPY returns over ~10 years. Parameters are point estimates subject to sampling uncertainty, particularly for the less-frequent regime states.
3. Transition matrix. Maximum likelihood estimate from observed sequence. Adjusted for macro stress via yield curve and recession probability using a heuristic stress factor (not empirically calibrated).
4. Recession overlay. Linear scaling from 0% drag at P(recession)=15% to -15% at P=70%. Applied to all regime means equally. The scaling coefficients are designed to produce reasonable adjustments, not fitted to historical data.
5. CAPE drift. E[r] ≈ 1/CAPE deviation from long-run mean, dampened 0.3x. The raw Shiller data extends to September 2023; we estimate the current CAPE (~38.0) by scaling for price change and approximate earnings growth, cross-validated against independent sources.
6. Simulation. 5,000 paths, 467 trading days. Base and enhanced runs use the same random seed for direct comparison.
7. Fan construction. Percentiles (10/25/50/75/90) computed cross-sectionally at each time step.

Limitations

• Regime-conditional returns are assumed normally distributed. Empirical returns exhibit fat tails, meaning the model understates extreme outcomes in both directions.
• CAPE is estimated from September 2023 raw data scaled by subsequent price changes and assumed 7% annual earnings growth. The estimate (~38.0) aligns with independent sources but the earnings growth assumption introduces uncertainty.
• The Neutral regime's -16.6% annualized return is heavily influenced by specific drawdown episodes (including COVID). A longer sample or different classification scheme could produce materially different Neutral parameters.
• The recession overlay, CAPE drift, and transition matrix adjustments all use heuristic coefficients rather than empirically optimized parameters. They are directionally defensible but not precisely calibrated.
• Cross-asset contagion, liquidity spirals, and nonlinear feedback effects are not modeled.

This is best interpreted as a conditional density estimate that integrates multiple macro signals into a coherent probability distribution — not as a point forecast or trading signal.