Monte Carlo Simulation in Retirement Planning

Most retirement projections use a single assumed rate of return — 7% per year, compounded neatly for 30 years. The result is a smooth, reassuring line on a graph that has almost nothing to do with how markets actually behave. Monte Carlo simulation replaces that single line with thousands of possible futures, each with a different sequence of returns, inflation, and market conditions. The result is not a prediction but a stress test: how does your spending plan hold up across the full range of what markets might do?

If you've read [Safe Withdrawal Rates](SafeWithdrawalRates) and understand why [Sequence of Returns Risk](SequenceOfReturnsRisk) makes the first decade of retirement so dangerous, Monte Carlo is the tool that ties those concepts together quantitatively.

Why Average Returns Aren't Enough

A spreadsheet projection using a 7% average return tells you one thing: the outcome if every year delivers exactly 7%. But no year delivers exactly 7%. Real returns vary wildly — up 28% one year, down 25% the next. When you're withdrawing money, the *order* of those returns matters as much as the average (see [Sequence of Returns Risk](SequenceOfReturnsRisk) for a detailed explanation with worked examples).

Consider two retirees with identical $1,000,000 portfolios, identical 5% withdrawal rates, and identical 7% average returns over 20 years. If bad returns come first, the portfolio may be exhausted by year 18. If good returns come first, it may grow to $1.8 million. Same average. Completely different outcomes. A deterministic spreadsheet cannot distinguish between these scenarios. Monte Carlo can.

How Monte Carlo Works

The simulation follows a straightforward process:

1. **Define inputs**: Expected return and volatility for each asset class, correlations between them, inflation assumptions, and withdrawal strategy

2. **Generate random paths**: For each simulation trial, randomly sample annual returns from the defined distributions, creating a unique 30-year (or longer) sequence

3. **Run the plan**: Apply your withdrawal strategy against each random path — deducting spending, adjusting for inflation, rebalancing

4. **Count survivors**: The percentage of trials where the portfolio lasts the full period is the "probability of success"

Most tools run 1,000 to 10,000 trials. Research on convergence shows that 1,000 iterations is adequate for typical planning, with results varying by less than 2% between runs. Only at the extreme tails (1st and 99th percentiles) does variation become meaningful, and even then 10,000 trials keeps it under 1.5%.

Return Modeling Approaches

|----------|-------------|----------|----------|

| **Parametric (lognormal)** | Generates returns from a bell-curve distribution using assumed mean and standard deviation | Unlimited scenarios; easy to adjust assumptions | Assumes returns follow a theoretical shape |

| **Historical bootstrapping** | Randomly draws actual past annual returns from the historical record | Preserves real-world distribution shape including fat tails | Limited to what has already happened; assumes future resembles past |

| **Regime-switching** | Uses Markov chains to alternate between bull and bear market states | Captures momentum and mean reversion; ~25% better accuracy than traditional MC | More complex; requires calibrating state transition probabilities |

Annual returns for a 60/40 portfolio pass the Shapiro-Wilk normality test, meaning the parametric approach is more defensible than critics suggest — at least for annual time steps. The real issue is not fat tails but *serial dependence*: real markets mean-revert over decades, while standard Monte Carlo assumes each year is independent. This causes Monte Carlo to actually **overstate** long-term risk. Its worst-case scenarios are worse than anything in the historical record.

What Monte Carlo Reveals About Spending Strategies

Monte Carlo's greatest value is comparing spending strategies head-to-head across identical market conditions. The table below summarizes the major approaches (see [Safe Withdrawal Rates](SafeWithdrawalRates) for deeper coverage of each):

|----------|-----------|---------------|------------------|-------------|

A Worked Example: David and Lisa Test Their Plan

David (65) and Lisa (63) have a $1,200,000 portfolio (60% stocks / 40% bonds) and want $54,000/year in portfolio withdrawals (4.5% initial rate). They expect $36,000/year combined Social Security starting at David's age 67. They run 10,000 Monte Carlo simulations over a 30-year horizon.

**Constant dollar approach** (inflation-adjusted $54,000):

- Probability of success: 82%

- Median ending portfolio: $1,450,000

- 10th percentile ending portfolio: $0 (depleted by year 26)

**Guardrails approach** (cut 10% if rate exceeds 5.4%, raise 10% if below 3.6%):

- Probability of success: 97%

- Median ending portfolio: $980,000

- Worst-case spending reduction: 28% (to $38,900/year — but Social Security provides $36,000 floor)

- Median spending: $52,000/year

The guardrails approach trades a modest average spending reduction for dramatically higher plan survival. And because David and Lisa have Social Security as a floor, even the worst-case spending cut still covers essentials. This is exactly the kind of insight Monte Carlo provides that deterministic planning cannot.

Interpreting the Results

The 95% Success Trap

A "95% probability of success" sounds reassuring. But the number deserves scrutiny from two directions:

**It may be too conservative.** Using the 4% rule, the median portfolio after 30 years is roughly **2.8 times the starting value**. In the typical scenario, retirees following this rule die with nearly three times their original savings — an enormous unspent surplus. Research comparing plans maintained at constant 95%, 70%, 50%, and even 20% probability found that median, minimum, and maximum spending levels were "actually quite consistent regardless of the probability of success used." The difference is primarily an income-versus-legacy tradeoff, not a safety-versus-ruin distinction.

**It may be misleading.** Success and failure are binary in the model — running out of money by $1 counts the same as running out by $500,000. For a couple receiving $4,500/month in Social Security with a $5,500/month spending target, portfolio depletion means an 18% lifestyle reduction. For someone entirely portfolio-dependent, it means 100% income loss. Same "failure rate," vastly different real-world consequences.

Reframing "Failure"

If you work with a financial planner and review your plan annually, a more useful interpretation is: the probability of success is really the **probability of NOT needing a spending adjustment at your next review**. A 70% success probability doesn't mean a 30% chance of ruin — it means a 30% chance you'll need to trim spending slightly at your next check-in, then reassess.

Spending Efficiency

How much of your accumulated wealth do you actually *use* during retirement? Blanchett's research on the "retirement spending smile" shows that real spending declines roughly 1% per year in inflation-adjusted terms during the middle decades of retirement, then rises at the end due to healthcare. This means constant-inflation-adjusted withdrawal assumptions overstate mid-retirement spending and understate late-retirement healthcare costs. Plans that model a realistic spending curve can support initial withdrawal rates approximately 1 percentage point higher.

Limitations and Caveats

Monte Carlo is a stress test, not a crystal ball. Important limitations:

**Input sensitivity.** Retirement outcomes are highly sensitive to capital market assumptions. Seemingly small changes — reducing expected equity returns from 8% to 6% — can mean the difference between projected comfort and projected ruin. Forward-looking estimates based on current valuations (CAPE, bond yields) outperform simple historical averages, but all are uncertain.

**U.S. historical bias.** Most tools calibrate to U.S. market history, which was exceptionally strong by international standards. Pfau's international research shows that the 4% rule would have failed in many developed countries.

**Behavioral gap.** Models assume retirees will rationally cut spending when dynamic rules dictate. Real humans resist lifestyle changes, maintain commitments, and experience psychological resistance to spending cuts — especially the 30-50% reductions that guardrails strategies can require in deep bear markets.

**Tax complexity.** Modeling federal taxes, state taxes, Social Security taxation thresholds, IRMAA brackets, and capital gains in different account types over 30 years compounds small errors into large ones. Research suggests that detailed tax assumptions are "inversely correlated to output accuracy" over long horizons. For how tax-aware withdrawal ordering interacts with these projections, see [Retirement Withdrawal Sequencing](RetirementWithdrawalSequencing).

**Healthcare costs.** Fidelity estimates $315,000+ in lifetime healthcare costs for a 65-year-old couple, with healthcare inflation running 5-7% versus 2-3% general CPI. Standard Monte Carlo typically models a single inflation rate, understating healthcare cost growth. See [Medicare Planning and Healthcare](MedicarePlanningAndHealthcare) for strategies to manage these costs.

Tools for Running Your Own Simulations

Several free tools make Monte Carlo accessible to individual investors:

|------|--------|-------------|----------|

What Inputs Matter Most

Focus your attention on these assumptions in order of impact:

1. **Expected returns** — the single most influential input. Use forward-looking estimates, not just historical averages

2. **Volatility** — a 30% loss requires a 43% gain to recover; this asymmetry drives sequence risk

3. **Correlations** — omitting these defaults to zero correlation, which dangerously overstates diversification. A simple two-asset model (stocks and bonds) is more honest than a complex multi-asset model with unmodeled correlations

4. **Time horizon** — safe withdrawal rates range from ~5.5% for 20 years to ~3.5% for 40 years

5. **Inflation** — regime-based modeling outperforms fixed assumptions but is not available in most consumer tools

Integrating With Other Retirement Decisions

Monte Carlo is most powerful when it evaluates interconnected decisions together:

- **Social Security timing**: Modern tools test every filing age combination against longevity and market uncertainty (see [Social Security Claiming Strategy](SocialSecurityClaimingStrategy))

- **Roth conversions**: Evaluate conversion amounts across tax brackets and their impact on future RMDs and IRMAA (see [Roth Conversion Strategy](RothConversionStrategy))

- **Withdrawal sequencing**: Determine the optimal order for tapping taxable, tax-deferred, and Roth accounts year by year (see [Retirement Withdrawal Sequencing](RetirementWithdrawalSequencing))

- **Income layering**: Build a multi-source income plan stress-tested across market conditions (see [Retirement Income Blueprint](RetirementIncomeBlueprint))

The value of Monte Carlo is not the specific probability it outputs. It is the ability to ask "what if" — what if markets crash in your first year, what if inflation spikes, what if you spend more than planned — and see the consequences play out across thousands of possible futures before committing to a strategy.